PS5 to officially do away with loading screens

Avatar image for rzxv04
rzxv04

2578

Forum Posts

0

Wiki Points

0

Followers

Reviews: 0

User Lists: 10

#201 rzxv04
Member since 2018 • 2578 Posts

@michaelmikado said:

@rzxv04: apples nvme isn’t faster than UFS 3.0. NVMe has higher theoretical bandwidth but in the real world UFS 3.0 transfer rates at are around 2.4GBps vs apples 1.8GBps. UFS 3.0 is also smaller, cheaper, and consumes less power. And in theory could sit directly on the interposer acting as an L4 cache. That said I’m also expecting DDR5 not 4 in these machines with a LP variant in later slim models.

So is UFS 3.0 the best price/performance/storage size currently? How would this compare to a theoretical x4 NVME on the PS5?

Avatar image for Xplode_games
Xplode_games

2540

Forum Posts

0

Wiki Points

0

Followers

Reviews: 0

User Lists: 0

#202 Xplode_games
Member since 2011 • 2540 Posts

They can do away with loading screens but they won't do away with loading. They can however greatly improve loading times and I am sure they will, especially with the rumored standard SSD.

Avatar image for michaelmikado
michaelmikado

406

Forum Posts

0

Wiki Points

0

Followers

Reviews: 0

User Lists: 5

#203 michaelmikado
Member since 2019 • 406 Posts

@ronvalencia said:
@michaelmikado said:

@ronvalencia:

I said theoretically as in it’s possible, not specifically for PS5 but for low end phones as UFS 3.0 is approaching LPDDR1 speeds. A 3.1 variant could likely allow low end phones to combine RAM and storage as a single unit. The point was to show how fast it is.

From Ryzen CPU's POV, RAM would be L4 while Intel Optane (8 GB/s to 33 GB/s) would be L5. https://www.zdnet.com/article/first-optane-performance-tests-show-benefits-and-limits-of-intels-nvdimms/

LPDDR defined by 16 bit or 32 bit bus.

What? No, it even specifies the link you provided how Optane would be used directly by the processor as it sits directly on the memory bus. That's kinda the whole point of NVDIMMS is that its now fast enough to serve as a RAM alternative.

Optane operates either as memory or in App Direct mode. Memory mode

. . . uses Optane DC to expand main memory capacity without persistence. It combines a Optane DC PMM with a conventional DRAM DIMM that serves as a direct-mapped cache for the Optane DC PMM. The CPU and operating system simply see a larger pool of main memory.

But ignoring that Optane chips you links costs somewhere between a car payment and a mortgage payment.

https://www.tomshardware.com/news/intel-optane-dimm-pricing-performance,39007.html

Intel Optane DIMM Pricing: $695 for 128GB, $2595 for 256GB, $7816 for 512GB

So even ignoring the feasiblity of pricing, the low-end variant comes in 16GB ($30-$40) and 32GB ($60-$80) flavors over an M2 interface. Ignoring their small size which would be better served by just putting more RAM in, the Sequential reads and writes are less than half of what UFS 3.0 would be. Even ignoring everything else you still have the issue of space, heat, power draw to contend with which is substantially higher than UFS 3.0.

Avatar image for michaelmikado
michaelmikado

406

Forum Posts

0

Wiki Points

0

Followers

Reviews: 0

User Lists: 5

#204 michaelmikado
Member since 2019 • 406 Posts

@rzxv04 said:
@michaelmikado said:

@rzxv04: apples nvme isn’t faster than UFS 3.0. NVMe has higher theoretical bandwidth but in the real world UFS 3.0 transfer rates at are around 2.4GBps vs apples 1.8GBps. UFS 3.0 is also smaller, cheaper, and consumes less power. And in theory could sit directly on the interposer acting as an L4 cache. That said I’m also expecting DDR5 not 4 in these machines with a LP variant in later slim models.

So is UFS 3.0 the best price/performance/storage size currently? How would this compare to a theoretical x4 NVME on the PS5?

NVM will have a theoretical performance of around 3.2GBps to UFS 3.0's 2.9GBps. Typical UFS 3.0 speeds average around 2.2-2.4GBps. I don't have an approximate price but its the next generation of the type of storage found in most smartphones so its pretty cheap. Also it can withstand much higher temperatures as well as part of the spec and has significantly less power draw and heat and size. Those last elements are vital for closed systems such as phones or consoles because it allows you to allot more of your "power budget" to other things. Realistically the ideal solution would be low power memory and storage and clock the APU where possible. A UFS 3.0 solution gives the best all around benefit and its the fastest type of NV storage I can think of that would fit the bill as well as meeting the criteria of being "unique" and not seen in PCs.

Avatar image for rzxv04
rzxv04

2578

Forum Posts

0

Wiki Points

0

Followers

Reviews: 0

User Lists: 10

#205 rzxv04
Member since 2018 • 2578 Posts

Loading: hides in Sony logo boot up / pre-game launch ; In game: none (ultimately depends on game design [PS4 GoW almost has none]).

Avatar image for ronvalencia
ronvalencia

29612

Forum Posts

0

Wiki Points

0

Followers

Reviews: 1

User Lists: 0

#206  Edited By ronvalencia
Member since 2008 • 29612 Posts

@michaelmikado said:
@ronvalencia said:
@michaelmikado said:

@ronvalencia:

I said theoretically as in it’s possible, not specifically for PS5 but for low end phones as UFS 3.0 is approaching LPDDR1 speeds. A 3.1 variant could likely allow low end phones to combine RAM and storage as a single unit. The point was to show how fast it is.

From Ryzen CPU's POV, RAM would be L4 while Intel Optane (8 GB/s to 33 GB/s) would be L5. https://www.zdnet.com/article/first-optane-performance-tests-show-benefits-and-limits-of-intels-nvdimms/

LPDDR defined by 16 bit or 32 bit bus.

What? No, it even specifies the link you provided how Optane would be used directly by the processor as it sits directly on the memory bus. That's kinda the whole point of NVDIMMS is that its now fast enough to serve as a RAM alternative.

Optane operates either as memory or in App Direct mode. Memory mode

. . . uses Optane DC to expand main memory capacity without persistence. It combines a Optane DC PMM with a conventional DRAM DIMM that serves as a direct-mapped cache for the Optane DC PMM. The CPU and operating system simply see a larger pool of main memory.

But ignoring that Optane chips you links costs somewhere between a car payment and a mortgage payment.

https://www.tomshardware.com/news/intel-optane-dimm-pricing-performance,39007.html

Intel Optane DIMM Pricing: $695 for 128GB, $2595 for 256GB, $7816 for 512GB

So even ignoring the feasiblity of pricing, the low-end variant comes in 16GB ($30-$40) and 32GB ($60-$80) flavors over an M2 interface. Ignoring their small size which would be better served by just putting more RAM in, the Sequential reads and writes are less than half of what UFS 3.0 would be. Even ignoring everything else you still have the issue of space, heat, power draw to contend with which is substantially higher than UFS 3.0.

High end mobile phones are not cheap.

LPDDR1-400 with 32bit bus yields 1.6 GB/s.

https://www.anandtech.com/show/14075/western-digital-develops-lowlatency-flash-to-compete-with-intel-optane

WD announces an alternative to Intel's Optane

https://pcper.com/2018/08/does-storemi-bring-amd-on-par-with-intel-optane-memory-caching/

AMD's StoreMI that combines RAM, SSD with HDD.

It has room to improve, but AMD StoreMI definitely closes a feature gap that the Ryzen platform had compared to Intel mainstream systems.

Competition in the PC industry drives the cost down.

Avatar image for rzxv04
rzxv04

2578

Forum Posts

0

Wiki Points

0

Followers

Reviews: 0

User Lists: 10

#207 rzxv04
Member since 2018 • 2578 Posts

@michaelmikado said:
@rzxv04 said:
@michaelmikado said:

@rzxv04: apples nvme isn’t faster than UFS 3.0. NVMe has higher theoretical bandwidth but in the real world UFS 3.0 transfer rates at are around 2.4GBps vs apples 1.8GBps. UFS 3.0 is also smaller, cheaper, and consumes less power. And in theory could sit directly on the interposer acting as an L4 cache. That said I’m also expecting DDR5 not 4 in these machines with a LP variant in later slim models.

So is UFS 3.0 the best price/performance/storage size currently? How would this compare to a theoretical x4 NVME on the PS5?

NVM will have a theoretical performance of around 3.2GBps to UFS 3.0's 2.9GBps. Typical UFS 3.0 speeds average around 2.2-2.4GBps. I don't have an approximate price but its the next generation of the type of storage found in most smartphones so its pretty cheap. Also it can withstand much higher temperatures as well as part of the spec and has significantly less power draw and heat and size. Those last elements are vital for closed systems such as phones or consoles because it allows you to allot more of your "power budget" to other things. Realistically the ideal solution would be low power memory and storage and clock the APU where possible. A UFS 3.0 solution gives the best all around benefit and its the fastest type of NV storage I can think of that would fit the bill as well as meeting the criteria of being "unique" and not seen in PCs.

So all that's left is for UFS 3.0 to come to mainstream? Will it have enough penetration by 2020 compared to NVME's?

Are there data for durability?

Can there be a NVME X5 and upwards to compete?

Thanks.

Avatar image for ronvalencia
ronvalencia

29612

Forum Posts

0

Wiki Points

0

Followers

Reviews: 1

User Lists: 0

#208  Edited By ronvalencia
Member since 2008 • 29612 Posts

@michaelmikado said:
@rzxv04 said:
@michaelmikado said:

@rzxv04: apples nvme isn’t faster than UFS 3.0. NVMe has higher theoretical bandwidth but in the real world UFS 3.0 transfer rates at are around 2.4GBps vs apples 1.8GBps. UFS 3.0 is also smaller, cheaper, and consumes less power. And in theory could sit directly on the interposer acting as an L4 cache. That said I’m also expecting DDR5 not 4 in these machines with a LP variant in later slim models.

So is UFS 3.0 the best price/performance/storage size currently? How would this compare to a theoretical x4 NVME on the PS5?

NVM will have a theoretical performance of around 3.2GBps to UFS 3.0's 2.9GBps. Typical UFS 3.0 speeds average around 2.2-2.4GBps. I don't have an approximate price but its the next generation of the type of storage found in most smartphones so its pretty cheap. Also it can withstand much higher temperatures as well as part of the spec and has significantly less power draw and heat and size. Those last elements are vital for closed systems such as phones or consoles because it allows you to allot more of your "power budget" to other things. Realistically the ideal solution would be low power memory and storage and clock the APU where possible. A UFS 3.0 solution gives the best all around benefit and its the fastest type of NV storage I can think of that would fit the bill as well as meeting the criteria of being "unique" and not seen in PCs.

AMD doesn't naively support mobile phone I/O e.g. display interface, UFS. UFS 2.0 is found in Snapdragon 835.

Loading Video...

https://en.wikipedia.org/wiki/Universal_Flash_Storage

Max bandwidth claims from UFS 3.0 is another bullshit from the mobile phone industry.

https://www.techspot.com/review/1651-nvme-ssd-performance-2018/

PC master race murders mobile industry's storage bandwidth bullshit claims.

The 970 Pro just barely edges out the WD Black in the read test but can't sustain its rated 3500MB/s speed for long. The 970 Evo had some troubles here and despite being rated at 3400MB/s, only achieved a max of about 3150MB/s before tapering off substantially. The Optane 800P modules are spot on to their advertised numbers and the 900P is well above its rated 2500MB/s.

Avatar image for michaelmikado
michaelmikado

406

Forum Posts

0

Wiki Points

0

Followers

Reviews: 0

User Lists: 5

#209 michaelmikado
Member since 2019 • 406 Posts

@rzxv04 said:

So all that's left is for UFS 3.0 to come to mainstream? Will it have enough penetration by 2020 compared to NVME's?

Are there data for durability?

Can there be a NVME X5 and upwards to compete?

Thanks.

UFS 3.0 really just released. It just hit mass production a few months ago by only a handful of manufacturers. It's the next generation of UFS and it's pretty much inevitable. That said, it's not a competitor to NVMe, one is for high-end PC and laptops while the other is for mobile devices where power and space are more of an issue. You may start to see it in low and mid-tier devices soon as it doesn't win on pure performance, rather its a well rounded solution. No real tests on durability because its only be out for a few years, but for what it's worth I have yet to personally see a UFS device fail.

As for NVMe, there will always be a better/faster solution but that's not the main objective. You have to balance your solution for a console and give the best performance/price/space/and low power consumption.

Avatar image for michaelmikado
michaelmikado

406

Forum Posts

0

Wiki Points

0

Followers

Reviews: 0

User Lists: 5

#210 michaelmikado
Member since 2019 • 406 Posts

@ronvalencia said:
@michaelmikado said:
@ronvalencia said:
@michaelmikado said:

@ronvalencia:

I said theoretically as in it’s possible, not specifically for PS5 but for low end phones as UFS 3.0 is approaching LPDDR1 speeds. A 3.1 variant could likely allow low end phones to combine RAM and storage as a single unit. The point was to show how fast it is.

From Ryzen CPU's POV, RAM would be L4 while Intel Optane (8 GB/s to 33 GB/s) would be L5. https://www.zdnet.com/article/first-optane-performance-tests-show-benefits-and-limits-of-intels-nvdimms/

LPDDR defined by 16 bit or 32 bit bus.

What? No, it even specifies the link you provided how Optane would be used directly by the processor as it sits directly on the memory bus. That's kinda the whole point of NVDIMMS is that its now fast enough to serve as a RAM alternative.

Optane operates either as memory or in App Direct mode. Memory mode

. . . uses Optane DC to expand main memory capacity without persistence. It combines a Optane DC PMM with a conventional DRAM DIMM that serves as a direct-mapped cache for the Optane DC PMM. The CPU and operating system simply see a larger pool of main memory.

But ignoring that Optane chips you links costs somewhere between a car payment and a mortgage payment.

https://www.tomshardware.com/news/intel-optane-dimm-pricing-performance,39007.html

Intel Optane DIMM Pricing: $695 for 128GB, $2595 for 256GB, $7816 for 512GB

So even ignoring the feasiblity of pricing, the low-end variant comes in 16GB ($30-$40) and 32GB ($60-$80) flavors over an M2 interface. Ignoring their small size which would be better served by just putting more RAM in, the Sequential reads and writes are less than half of what UFS 3.0 would be. Even ignoring everything else you still have the issue of space, heat, power draw to contend with which is substantially higher than UFS 3.0.

High end mobile phones are not cheap.

LPDDR1-400 with 32bit bus yields 1.6 GB/s.

https://www.anandtech.com/show/14075/western-digital-develops-lowlatency-flash-to-compete-with-intel-optane

WD announces an alternative to Intel's Optane

https://pcper.com/2018/08/does-storemi-bring-amd-on-par-with-intel-optane-memory-caching/

AMD's StoreMI that combines RAM, SSD with HDD.

It has room to improve, but AMD StoreMI definitely closes a feature gap that the Ryzen platform had compared to Intel mainstream systems.

Competition in the PC industry drives the cost down.

I'm going to respond to this because they are somewhat related to the PS5.

From the first article:

The manufacturer admits that its LLF is a customized device that is very fast because it is tailored for performance. The memory will cost 10 times less than DRAM, but 20 times more than 3D NAND (at least based on today’s estimates) in terms of per-GB prices, so it will likely be used only by select applications aimed at datacenters or high-end workstations, similar to where Optane and Z-NAND is today.

Again, this is not a balances solution, there comes a point where performance is "good enough" for the price you pay which a next gen flash storage solution is that.

As for the second article it's a software solution not an actual physical solution.

The PS5 is rumored to have similar software which sees all the RAM and cache as a single pool and dynamically allocates workloads where needed. The link is somewhere out there, I'm not interested in exerting the effort to find it.

Avatar image for ronvalencia
ronvalencia

29612

Forum Posts

0

Wiki Points

0

Followers

Reviews: 1

User Lists: 0

#211  Edited By ronvalencia
Member since 2008 • 29612 Posts

@michaelmikado said:
@ronvalencia said:
@michaelmikado said:
@ronvalencia said:

From Ryzen CPU's POV, RAM would be L4 while Intel Optane (8 GB/s to 33 GB/s) would be L5. https://www.zdnet.com/article/first-optane-performance-tests-show-benefits-and-limits-of-intels-nvdimms/

LPDDR defined by 16 bit or 32 bit bus.

What? No, it even specifies the link you provided how Optane would be used directly by the processor as it sits directly on the memory bus. That's kinda the whole point of NVDIMMS is that its now fast enough to serve as a RAM alternative.

Optane operates either as memory or in App Direct mode. Memory mode

. . . uses Optane DC to expand main memory capacity without persistence. It combines a Optane DC PMM with a conventional DRAM DIMM that serves as a direct-mapped cache for the Optane DC PMM. The CPU and operating system simply see a larger pool of main memory.

But ignoring that Optane chips you links costs somewhere between a car payment and a mortgage payment.

https://www.tomshardware.com/news/intel-optane-dimm-pricing-performance,39007.html

Intel Optane DIMM Pricing: $695 for 128GB, $2595 for 256GB, $7816 for 512GB

So even ignoring the feasiblity of pricing, the low-end variant comes in 16GB ($30-$40) and 32GB ($60-$80) flavors over an M2 interface. Ignoring their small size which would be better served by just putting more RAM in, the Sequential reads and writes are less than half of what UFS 3.0 would be. Even ignoring everything else you still have the issue of space, heat, power draw to contend with which is substantially higher than UFS 3.0.

High end mobile phones are not cheap.

LPDDR1-400 with 32bit bus yields 1.6 GB/s.

https://www.anandtech.com/show/14075/western-digital-develops-lowlatency-flash-to-compete-with-intel-optane

WD announces an alternative to Intel's Optane

https://pcper.com/2018/08/does-storemi-bring-amd-on-par-with-intel-optane-memory-caching/

AMD's StoreMI that combines RAM, SSD with HDD.

It has room to improve, but AMD StoreMI definitely closes a feature gap that the Ryzen platform had compared to Intel mainstream systems.

Competition in the PC industry drives the cost down.

I'm going to respond to this because they are somewhat related to the PS5.

From the first article:

The manufacturer admits that its LLF is a customized device that is very fast because it is tailored for performance. The memory will cost 10 times less than DRAM, but 20 times more than 3D NAND (at least based on today’s estimates) in terms of per-GB prices, so it will likely be used only by select applications aimed at datacenters or high-end workstations, similar to where Optane and Z-NAND is today.

Again, this is not a balances solution, there comes a point where performance is "good enough" for the price you pay which a next gen flash storage solution is that.

As for the second article it's a software solution not an actual physical solution.

The PS5 is rumored to have similar software which sees all the RAM and cache as a single pool and dynamically allocates workloads where needed. The link is somewhere out there, I'm not interested in exerting the effort to find it.

There's a high probability it's semi-custom storage acceleration stack from AMD which evolved from StoreMI. Zen v2 comes with new chip-sets and PCI-E version 4.0 infrastructure.

PS4's CPU-to-GPU links follows PCI-E +2.0 limits i.e. PS4's 10 GB/s full duplex (20 GB/s total) vs PCI-E 2.0 16X 8GB/s full duplex (16 GB/s total)

XBO's CPU-to-GPU links follows PCI-E 3.0 16X limits i.e. ~33 GB/s total.

Avatar image for rzxv04
rzxv04

2578

Forum Posts

0

Wiki Points

0

Followers

Reviews: 0

User Lists: 10

#212  Edited By rzxv04
Member since 2018 • 2578 Posts

@michaelmikado said:
@rzxv04 said:

So all that's left is for UFS 3.0 to come to mainstream? Will it have enough penetration by 2020 compared to NVME's?

Are there data for durability?

Can there be a NVME X5 and upwards to compete?

Thanks.

UFS 3.0 really just released. It just hit mass production a few months ago by only a handful of manufacturers. It's the next generation of UFS and it's pretty much inevitable. That said, it's not a competitor to NVMe, one is for high-end PC and laptops while the other is for mobile devices where power and space are more of an issue. You may start to see it in low and mid-tier devices soon as it doesn't win on pure performance, rather its a well rounded solution. No real tests on durability because its only be out for a few years, but for what it's worth I have yet to personally see a UFS device fail.

As for NVMe, there will always be a better/faster solution but that's not the main objective. You have to balance your solution for a console and give the best performance/price/space/and low power consumption.

So in short, NVME is currently a more likely candidate for the upcoming consoles in pursuit of "no load screens" barring mid-gen iterations due to commonality with PC hardware of AMD?

I was thinking that NVME is becoming quite cheap on the PC if you visit sites like slickdeals plus maybe AMD and Sony can give it more "lanes" for the PS5 but not much change the NVME.

At the end though, I think even if the bare minimum for consoles are leveraged to NVME or even Sata 3.0, Game design will take care of the creating better illusions of loads.

Avatar image for michaelmikado
michaelmikado

406

Forum Posts

0

Wiki Points

0

Followers

Reviews: 0

User Lists: 5

#213 michaelmikado
Member since 2019 • 406 Posts

@rzxv04:

No it would be the opposite, it’s a storage media the idea that the processor would be unable to accept a different storage media that’s literally a semi custom design any is simply untrue.

Further any high end SSD is going to be low volume on the PC side. By contrast UFS is ubiquitous to cellphones, Chromebooks etc. as well so the price dramatically drops over time, again ignoring the other benefits of power consumption and space it is a better option at the targeted limitations of a console. There simply isn’t any benefit to use nvme in a console vs UFS 3.0 which is technically faster than nvme on paper.

Avatar image for rzxv04
rzxv04

2578

Forum Posts

0

Wiki Points

0

Followers

Reviews: 0

User Lists: 10

#214 rzxv04
Member since 2018 • 2578 Posts

@michaelmikado said:

@rzxv04:

No it would be the opposite, it’s a storage media the idea that the processor would be unable to accept a different storage media that’s literally a semi custom design any is simply untrue.

Further any high end SSD is going to be low volume on the PC side. By contrast UFS is ubiquitous to cellphones, Chromebooks etc. as well so the price dramatically drops over time, again ignoring the other benefits of power consumption and space it is a better option at the targeted limitations of a console. There simply isn’t any benefit to use nvme in a console vs UFS 3.0 which is technically faster than nvme on paper.

I see. So the question would be if it's gonna hit that sweet spot in availability and pricing for the release of the PS5?

Avatar image for michaelmikado
michaelmikado

406

Forum Posts

0

Wiki Points

0

Followers

Reviews: 0

User Lists: 5

#215 michaelmikado
Member since 2019 • 406 Posts

@rzxv04 said:
@michaelmikado said:

@rzxv04:

No it would be the opposite, it’s a storage media the idea that the processor would be unable to accept a different storage media that’s literally a semi custom design any is simply untrue.

Further any high end SSD is going to be low volume on the PC side. By contrast UFS is ubiquitous to cellphones, Chromebooks etc. as well so the price dramatically drops over time, again ignoring the other benefits of power consumption and space it is a better option at the targeted limitations of a console. There simply isn’t any benefit to use nvme in a console vs UFS 3.0 which is technically faster than nvme on paper.

I see. So the question would be if it's gonna hit that sweet spot in availability and pricing for the release of the PS5?

Not really, its available now and less than NVMe drives. For context sake we know the OnePlus 7 Pro has 128GB vs 256GB has a price difference of about $30 bucks. It would be a fair assumption to think the 128GB drive is $30 and the 256GB is $60. By contrast a comparable nvme drive is going to run you more than twice that on Newegg. https://www.newegg.com/samsung-960-evo-250gb/p/N82E16820147593

I'm not saying this is definitely it, but it makes the most sense and they could even have a more custom variant with higher performance.

Avatar image for rzxv04
rzxv04

2578

Forum Posts

0

Wiki Points

0

Followers

Reviews: 0

User Lists: 10

#216 rzxv04
Member since 2018 • 2578 Posts

@michaelmikado said:
@rzxv04 said:
@michaelmikado said:

@rzxv04:

No it would be the opposite, it’s a storage media the idea that the processor would be unable to accept a different storage media that’s literally a semi custom design any is simply untrue.

Further any high end SSD is going to be low volume on the PC side. By contrast UFS is ubiquitous to cellphones, Chromebooks etc. as well so the price dramatically drops over time, again ignoring the other benefits of power consumption and space it is a better option at the targeted limitations of a console. There simply isn’t any benefit to use nvme in a console vs UFS 3.0 which is technically faster than nvme on paper.

I see. So the question would be if it's gonna hit that sweet spot in availability and pricing for the release of the PS5?

Not really, its available now and less than NVMe drives. For context sake we know the OnePlus 7 Pro has 128GB vs 256GB has a price difference of about $30 bucks. It would be a fair assumption to think the 128GB drive is $30 and the 256GB is $60. By contrast a comparable nvme drive is going to run you more than twice that on Newegg. https://www.newegg.com/samsung-960-evo-250gb/p/N82E16820147593

I'm not saying this is definitely it, but it makes the most sense and they could even have a more custom variant with higher performance.

Thanks.

I can't seem to understand this guy's presentation. What exactly does he mean that UFS 3.0 isn't as fast?

https://www.youtube.com/watch?v=TaKC1qIlo9Y&t=572s

NVME can seem to come down in price through lots of sales the past few months about $ 100 / TB:

https://slickdeals.net/deals/ssd/

https://www.microcenter.com/product/600422/1TB_SSD_3D_NAND_M2_2280_PCIe_NVMe_30_x4_Internal_Solid_State_Drive

I imagine bulk purchases by Sony and MS would be much cheaper.

How inferior is NVME x2 compared to UFS 3.0?

Avatar image for ronvalencia
ronvalencia

29612

Forum Posts

0

Wiki Points

0

Followers

Reviews: 1

User Lists: 0

#217  Edited By ronvalencia
Member since 2008 • 29612 Posts

@rzxv04 said:
@michaelmikado said:
@rzxv04 said:
@michaelmikado said:

@rzxv04:

No it would be the opposite, it’s a storage media the idea that the processor would be unable to accept a different storage media that’s literally a semi custom design any is simply untrue.

Further any high end SSD is going to be low volume on the PC side. By contrast UFS is ubiquitous to cellphones, Chromebooks etc. as well so the price dramatically drops over time, again ignoring the other benefits of power consumption and space it is a better option at the targeted limitations of a console. There simply isn’t any benefit to use nvme in a console vs UFS 3.0 which is technically faster than nvme on paper.

I see. So the question would be if it's gonna hit that sweet spot in availability and pricing for the release of the PS5?

Not really, its available now and less than NVMe drives. For context sake we know the OnePlus 7 Pro has 128GB vs 256GB has a price difference of about $30 bucks. It would be a fair assumption to think the 128GB drive is $30 and the 256GB is $60. By contrast a comparable nvme drive is going to run you more than twice that on Newegg. https://www.newegg.com/samsung-960-evo-250gb/p/N82E16820147593

I'm not saying this is definitely it, but it makes the most sense and they could even have a more custom variant with higher performance.

Thanks.

I can't seem to understand this guy's presentation. What exactly does he mean that UFS 3.0 isn't as fast?

https://www.youtube.com/watch?v=TaKC1qIlo9Y&t=572s

NVME can seem to come down in price through lots of sales the past few months about $ 100 / TB:

https://slickdeals.net/deals/ssd/

https://www.microcenter.com/product/600422/1TB_SSD_3D_NAND_M2_2280_PCIe_NVMe_30_x4_Internal_Solid_State_Drive

I imagine bulk purchases by Sony and MS would be much cheaper.

How inferior is NVME x2 compared to UFS 3.0?

UFS 3.0 is slow when compared to PC's good branded SSD with SATA III or NVMe. Qualcomm's Snapdragon 855 SoC's I/O could be an issue. UFS project is mainly Qualcomm's initiative.

https://www.tweaktown.com/reviews/8176/samsung-pm961-1tb-2-nvme-pcie-ssd-review/index.html

Samsung PM961 review which is nearly comparable to Samsung 960 evo

https://pcpartpicker.com/product/k64NnQ/samsung-pm961-256gb-m2-2280-solid-state-drive-mzvlw256hehp-00000

Samsung PM961 256GB's $49 retail price.

Avatar image for michaelmikado
michaelmikado

406

Forum Posts

0

Wiki Points

0

Followers

Reviews: 0

User Lists: 5

#218 michaelmikado
Member since 2019 • 406 Posts

@rzxv04 said:
@michaelmikado said:
@rzxv04 said:
@michaelmikado said:

@rzxv04:

No it would be the opposite, it’s a storage media the idea that the processor would be unable to accept a different storage media that’s literally a semi custom design any is simply untrue.

Further any high end SSD is going to be low volume on the PC side. By contrast UFS is ubiquitous to cellphones, Chromebooks etc. as well so the price dramatically drops over time, again ignoring the other benefits of power consumption and space it is a better option at the targeted limitations of a console. There simply isn’t any benefit to use nvme in a console vs UFS 3.0 which is technically faster than nvme on paper.

I see. So the question would be if it's gonna hit that sweet spot in availability and pricing for the release of the PS5?

Not really, its available now and less than NVMe drives. For context sake we know the OnePlus 7 Pro has 128GB vs 256GB has a price difference of about $30 bucks. It would be a fair assumption to think the 128GB drive is $30 and the 256GB is $60. By contrast a comparable nvme drive is going to run you more than twice that on Newegg. https://www.newegg.com/samsung-960-evo-250gb/p/N82E16820147593

I'm not saying this is definitely it, but it makes the most sense and they could even have a more custom variant with higher performance.

Thanks.

I can't seem to understand this guy's presentation. What exactly does he mean that UFS 3.0 isn't as fast?

https://www.youtube.com/watch?v=TaKC1qIlo9Y&t=572s

NVME can seem to come down in price through lots of sales the past few months about $ 100 / TB:

https://slickdeals.net/deals/ssd/

https://www.microcenter.com/product/600422/1TB_SSD_3D_NAND_M2_2280_PCIe_NVMe_30_x4_Internal_Solid_State_Drive

I imagine bulk purchases by Sony and MS would be much cheaper.

How inferior is NVME x2 compared to UFS 3.0?

He is referencing sequential reads vs random reads. NVMe excels at random reads. UFS excels at sequential. Even at the lowest real world estimates for UFS 3.0 it out paces the highest theoretical performance for NVMe x2 at 1.6-1.8GBps in sequential reads but UFS suffers in random reads. Random reads really benefit exactly from what it sounds like. Unpredictable activity such as launching apps by the end user while sequential reads would be something like 4K video where the information is kept in blocks located close together on the storage media. Again, in a console you do not have a bunch of apps opening. The console has one primary function so it will be relatively easy to manage memory. Even in open world games there is only a finite amount you need to swap into memory at high speed and any load times that would occur would be so minimal that a developer could mask them easily even for something for fast travel. Basically it would enable load times in the single digits if even that for far cheaper and less cost all around and more wattage can be allocated to the GPU or RAM which is vitally important. You wouldn't want an SSD to suck up a significant amount of cost, power, or space at the expense of everything else just to get a second or two less of already low load times. That's what consoles are trade offs because there comes a power where solutions are "good enough" and balance out for the best for the consoles.

Avatar image for ronvalencia
ronvalencia

29612

Forum Posts

0

Wiki Points

0

Followers

Reviews: 1

User Lists: 0

#219  Edited By ronvalencia
Member since 2008 • 29612 Posts

@michaelmikado said:
@rzxv04 said:
@michaelmikado said:
@rzxv04 said:

I see. So the question would be if it's gonna hit that sweet spot in availability and pricing for the release of the PS5?

Not really, its available now and less than NVMe drives. For context sake we know the OnePlus 7 Pro has 128GB vs 256GB has a price difference of about $30 bucks. It would be a fair assumption to think the 128GB drive is $30 and the 256GB is $60. By contrast a comparable nvme drive is going to run you more than twice that on Newegg. https://www.newegg.com/samsung-960-evo-250gb/p/N82E16820147593

I'm not saying this is definitely it, but it makes the most sense and they could even have a more custom variant with higher performance.

Thanks.

I can't seem to understand this guy's presentation. What exactly does he mean that UFS 3.0 isn't as fast?

https://www.youtube.com/watch?v=TaKC1qIlo9Y&t=572s

NVME can seem to come down in price through lots of sales the past few months about $ 100 / TB:

https://slickdeals.net/deals/ssd/

https://www.microcenter.com/product/600422/1TB_SSD_3D_NAND_M2_2280_PCIe_NVMe_30_x4_Internal_Solid_State_Drive

I imagine bulk purchases by Sony and MS would be much cheaper.

How inferior is NVME x2 compared to UFS 3.0?

He is referencing sequential reads vs random reads. NVMe excels at random reads. UFS excels at sequential. Even at the lowest real world estimates for UFS 3.0 it out paces the highest theoretical performance for NVMe x2 at 1.6-1.8GBps in sequential reads but UFS suffers in random reads. Random reads really benefit exactly from what it sounds like. Unpredictable activity such as launching apps by the end user while sequential reads would be something like 4K video where the information is kept in blocks located close together on the storage media. Again, in a console you do not have a bunch of apps opening. The console has one primary function so it will be relatively easy to manage memory. Even in open world games there is only a finite amount you need to swap into memory at high speed and any load times that would occur would be so minimal that a developer could mask them easily even for something for fast travel. Basically it would enable load times in the single digits if even that for far cheaper and less cost all around and more wattage can be allocated to the GPU or RAM which is vitally important. You wouldn't want an SSD to suck up a significant amount of cost, power, or space at the expense of everything else just to get a second or two less of already low load times. That's what consoles are trade offs because there comes a power where solutions are "good enough" and balance out for the best for the consoles.

NVMe excels in both random and sequential

https://www.tweaktown.com/reviews/8176/samsung-pm961-1tb-2-nvme-pcie-ssd-review/index4.html

Sequential read reached ~3.1 GB/sbenchmarked (not theoretical) for Samsung PM961 nvme.

Highest theoretical performance is useless. Ryzen's results are slightly less than Intel's.

Sony has used laptop 2.5 inch HDDs in PS4s.

Avatar image for rzxv04
rzxv04

2578

Forum Posts

0

Wiki Points

0

Followers

Reviews: 0

User Lists: 10

#220 rzxv04
Member since 2018 • 2578 Posts

Thanks, you two.

So I guess it'll depend on better price and mass availability.

NVMEs seem to be dropping fast and if the trend continues may have a price advantage from $ 60 256GB UFS 3.0 but NVME has worse power consumption and space (by how much?).

That PM961 seems expensive.

Would Sony have to pay extra for paying for UFS 3.0 controllers?

Not sure how practical this comparison is 7 Pro vs XS Max

https://www.youtube.com/watch?v=Rs2bmYhW738

I'm surprised how many times the android was faster and from what I know it's bottom tier in terms of dev optimization for majority of apps. Maybe if Apple adopts UFS 3.0 we'll be able to see it really shine because devs will probably optimize more for apple apps.

Avatar image for ronvalencia
ronvalencia

29612

Forum Posts

0

Wiki Points

0

Followers

Reviews: 1

User Lists: 0

#221  Edited By ronvalencia
Member since 2008 • 29612 Posts

@rzxv04 said:

Thanks, you two.

So I guess it'll depend on better price and mass availability.

NVMEs seem to be dropping fast and if the trend continues may have a price advantage from $ 60 256GB UFS 3.0 but NVME has worse power consumption and space (by how much?).

That PM961 seems expensive.

Would Sony have to pay extra for paying for UFS 3.0 controllers?

Not sure how practical this comparison is 7 Pro vs XS Max

https://www.youtube.com/watch?v=Rs2bmYhW738

I'm surprised how many times the android was faster and from what I know it's bottom tier in terms of dev optimization for majority of apps. Maybe if Apple adopts UFS 3.0 we'll be able to see it really shine because devs will probably optimize more for apple apps.

Sony has selected laptop 2.5 inch HDDs with their PS4s.

https://pcpartpicker.com/product/xpYLrH/crucial-p1-500gb-m2-2280-solid-state-drive-ct500p1ssd8

Crucial P1 with 500 GB M.2-2280 for $59.99 retail

http://www.relaxedtech.com/reviews/crucial/p1/

Crucial P1 reached 1,817 MB/s sequential read.

Sony has tolerated ~2.5 watts laptop 2.5 inch HDDs.

15 seconds to 0.8 seconds would need about 18X improvement.

60 GB/s 5400 rpm HDD x 18 = 1,1125 MB/s

100 GB/s 5400 rpm HDD x 18 = 1,800 MB/s

Sony seems to have selected middle tier NVMe class storage solution

Avatar image for Jag85
Jag85

20701

Forum Posts

0

Wiki Points

0

Followers

Reviews: 219

User Lists: 0

#222 Jag85
Member since 2005 • 20701 Posts

@R4gn4r0k said:
@BassMan said:

Even playing games with an NVMe SSD and a powerful CPU on PC does not get rid of load times. So, he is full of shit. Assets and data need to be loaded and processed. That shit takes time and will not be eliminated completely.

A lot of casual gamers don't really see the reality, and people can't always be that informed. Think back to when you were a more casual gamer. With less knowledge about these technical details.

So if sony keeps throwing around buzz words like:

- 8K

- 120fps

- SSD

- no load times

Those things just get stuck in people's head and they create hype, eventhough they will all turn out to be lies (and have turned out to be lies so many times in the past) Sony sells consoles this way.

That is how business is done.

Let's not forget buzzwords like:

- Power of PlayStation

- Emotion Engine

- Cell

- 4 The Gamers

Avatar image for rzxv04
rzxv04

2578

Forum Posts

0

Wiki Points

0

Followers

Reviews: 0

User Lists: 10

#223 rzxv04
Member since 2018 • 2578 Posts

@ronvalencia said:
@rzxv04 said:

Thanks, you two.

So I guess it'll depend on better price and mass availability.

NVMEs seem to be dropping fast and if the trend continues may have a price advantage from $ 60 256GB UFS 3.0 but NVME has worse power consumption and space (by how much?).

That PM961 seems expensive.

Would Sony have to pay extra for paying for UFS 3.0 controllers?

Not sure how practical this comparison is 7 Pro vs XS Max

https://www.youtube.com/watch?v=Rs2bmYhW738

I'm surprised how many times the android was faster and from what I know it's bottom tier in terms of dev optimization for majority of apps. Maybe if Apple adopts UFS 3.0 we'll be able to see it really shine because devs will probably optimize more for apple apps.

Sony has selected laptop 2.5 inch HDDs with their PS4s.

https://pcpartpicker.com/product/xpYLrH/crucial-p1-500gb-m2-2280-solid-state-drive-ct500p1ssd8

Crucial P1 with 500 GB M.2-2280 for $59.99 retail

http://www.relaxedtech.com/reviews/crucial/p1/

Crucial P1 reached 1,817 MB/s sequential read.

Sony has tolerated ~2.5 watts laptop 2.5 inch HDDs.

15 seconds to 0.8 seconds would need about 18X improvement.

60 GB/s 5400 rpm HDD x 18 = 1,1125 MB/s

100 GB/s 5400 rpm HDD x 18 = 1,800 MB/s

Sony seems to have selected middle tier NVMe class storage solution

Thanks.

What do you think Cerny exactly meant by ".. but Cerny claims that it has a raw bandwidth higher than any SSD available for PCs. That’s not all. “The raw read speed is important,“ Cerny says, “but so are the details of the I/O [input-output] mechanisms and the software stack that we put on top of them. I got a PlayStation 4 Pro and then I put in a SSD that cost as much as the PlayStation 4 Pro—it might be one-third faster." As opposed to 19 times faster for the next-gen console, judging from the fast-travel demo."

Could it just be a PCIE 4.0 thing or do you think it's pure BS for cerny to say that.

Avatar image for ronvalencia
ronvalencia

29612

Forum Posts

0

Wiki Points

0

Followers

Reviews: 1

User Lists: 0

#224  Edited By ronvalencia
Member since 2008 • 29612 Posts

@rzxv04 said:
@ronvalencia said:
@rzxv04 said:

Thanks, you two.

So I guess it'll depend on better price and mass availability.

NVMEs seem to be dropping fast and if the trend continues may have a price advantage from $ 60 256GB UFS 3.0 but NVME has worse power consumption and space (by how much?).

That PM961 seems expensive.

Would Sony have to pay extra for paying for UFS 3.0 controllers?

Not sure how practical this comparison is 7 Pro vs XS Max

https://www.youtube.com/watch?v=Rs2bmYhW738

I'm surprised how many times the android was faster and from what I know it's bottom tier in terms of dev optimization for majority of apps. Maybe if Apple adopts UFS 3.0 we'll be able to see it really shine because devs will probably optimize more for apple apps.

Sony has selected laptop 2.5 inch HDDs with their PS4s.

https://pcpartpicker.com/product/xpYLrH/crucial-p1-500gb-m2-2280-solid-state-drive-ct500p1ssd8

Crucial P1 with 500 GB M.2-2280 for $59.99 retail

http://www.relaxedtech.com/reviews/crucial/p1/

Crucial P1 reached 1,817 MB/s sequential read.

Sony has tolerated ~2.5 watts laptop 2.5 inch HDDs.

15 seconds to 0.8 seconds would need about 18X improvement.

60 GB/s 5400 rpm HDD x 18 = 1,1125 MB/s

100 GB/s 5400 rpm HDD x 18 = 1,800 MB/s

Sony seems to have selected middle tier NVMe class storage solution

Thanks.

What do you think Cerny exactly meant by ".. but Cerny claims that it has a raw bandwidth higher than any SSD available for PCs. That’s not all. “The raw read speed is important,“ Cerny says, “but so are the details of the I/O [input-output] mechanisms and the software stack that we put on top of them. I got a PlayStation 4 Pro and then I put in a SSD that cost as much as the PlayStation 4 Pro—it might be one-third faster." As opposed to 19 times faster for the next-gen console, judging from the fast-travel demo."

Could it just be a PCIE 4.0 thing or do you think it's pure BS for cerny to say that.

1. SATA III would bottleneck any good SSD. "One-third faster" with SSD indicates other bottlenecks with PS4 Pro's storage controllers.

2. Define which SSD? SATA SSD or NVMe SSD?

Avatar image for rzxv04
rzxv04

2578

Forum Posts

0

Wiki Points

0

Followers

Reviews: 0

User Lists: 10

#225 rzxv04
Member since 2018 • 2578 Posts

@ronvalencia said:
@rzxv04 said:
@ronvalencia said:
@rzxv04 said:

Thanks, you two.

So I guess it'll depend on better price and mass availability.

NVMEs seem to be dropping fast and if the trend continues may have a price advantage from $ 60 256GB UFS 3.0 but NVME has worse power consumption and space (by how much?).

That PM961 seems expensive.

Would Sony have to pay extra for paying for UFS 3.0 controllers?

Not sure how practical this comparison is 7 Pro vs XS Max

https://www.youtube.com/watch?v=Rs2bmYhW738

I'm surprised how many times the android was faster and from what I know it's bottom tier in terms of dev optimization for majority of apps. Maybe if Apple adopts UFS 3.0 we'll be able to see it really shine because devs will probably optimize more for apple apps.

Sony has selected laptop 2.5 inch HDDs with their PS4s.

https://pcpartpicker.com/product/xpYLrH/crucial-p1-500gb-m2-2280-solid-state-drive-ct500p1ssd8

Crucial P1 with 500 GB M.2-2280 for $59.99 retail

http://www.relaxedtech.com/reviews/crucial/p1/

Crucial P1 reached 1,817 MB/s sequential read.

Sony has tolerated ~2.5 watts laptop 2.5 inch HDDs.

15 seconds to 0.8 seconds would need about 18X improvement.

60 GB/s 5400 rpm HDD x 18 = 1,1125 MB/s

100 GB/s 5400 rpm HDD x 18 = 1,800 MB/s

Sony seems to have selected middle tier NVMe class storage solution

Thanks.

What do you think Cerny exactly meant by ".. but Cerny claims that it has a raw bandwidth higher than any SSD available for PCs. That’s not all. “The raw read speed is important,“ Cerny says, “but so are the details of the I/O [input-output] mechanisms and the software stack that we put on top of them. I got a PlayStation 4 Pro and then I put in a SSD that cost as much as the PlayStation 4 Pro—it might be one-third faster." As opposed to 19 times faster for the next-gen console, judging from the fast-travel demo."

Could it just be a PCIE 4.0 thing or do you think it's pure BS for cerny to say that.

1. SATA III would bottleneck any good SSD. "One-third faster" with SSD indicates other bottlenecks with PS4 Pro's storage controllers.

2. Define which SSD? SATA SSD or NVMe SSD?

1. Makes sense.

2. Said "any ssd available for PC".

Would it be possible for them to currently have the highest bandwidth for a current mass market ssd of but somehow slower in other metrics? Something like wide but slower?

Avatar image for michaelmikado
michaelmikado

406

Forum Posts

0

Wiki Points

0

Followers

Reviews: 0

User Lists: 5

#226  Edited By michaelmikado
Member since 2019 • 406 Posts

@ronvalencia said:
@michaelmikado said:
@rzxv04 said:
@michaelmikado said:

Not really, its available now and less than NVMe drives. For context sake we know the OnePlus 7 Pro has 128GB vs 256GB has a price difference of about $30 bucks. It would be a fair assumption to think the 128GB drive is $30 and the 256GB is $60. By contrast a comparable nvme drive is going to run you more than twice that on Newegg. https://www.newegg.com/samsung-960-evo-250gb/p/N82E16820147593

I'm not saying this is definitely it, but it makes the most sense and they could even have a more custom variant with higher performance.

Thanks.

I can't seem to understand this guy's presentation. What exactly does he mean that UFS 3.0 isn't as fast?

https://www.youtube.com/watch?v=TaKC1qIlo9Y&t=572s

NVME can seem to come down in price through lots of sales the past few months about $ 100 / TB:

https://slickdeals.net/deals/ssd/

https://www.microcenter.com/product/600422/1TB_SSD_3D_NAND_M2_2280_PCIe_NVMe_30_x4_Internal_Solid_State_Drive

I imagine bulk purchases by Sony and MS would be much cheaper.

How inferior is NVME x2 compared to UFS 3.0?

He is referencing sequential reads vs random reads. NVMe excels at random reads. UFS excels at sequential. Even at the lowest real world estimates for UFS 3.0 it out paces the highest theoretical performance for NVMe x2 at 1.6-1.8GBps in sequential reads but UFS suffers in random reads. Random reads really benefit exactly from what it sounds like. Unpredictable activity such as launching apps by the end user while sequential reads would be something like 4K video where the information is kept in blocks located close together on the storage media. Again, in a console you do not have a bunch of apps opening. The console has one primary function so it will be relatively easy to manage memory. Even in open world games there is only a finite amount you need to swap into memory at high speed and any load times that would occur would be so minimal that a developer could mask them easily even for something for fast travel. Basically it would enable load times in the single digits if even that for far cheaper and less cost all around and more wattage can be allocated to the GPU or RAM which is vitally important. You wouldn't want an SSD to suck up a significant amount of cost, power, or space at the expense of everything else just to get a second or two less of already low load times. That's what consoles are trade offs because there comes a power where solutions are "good enough" and balance out for the best for the consoles.

NVMe excels in both random and sequential

https://www.tweaktown.com/reviews/8176/samsung-pm961-1tb-2-nvme-pcie-ssd-review/index4.html

Sequential read reached ~3.1 GB/sbenchmarked (not theoretical) for Samsung PM961 nvme.

Highest theoretical performance is useless. Ryzen's results are slightly less than Intel's.

Sony has used laptop 2.5 inch HDDs in PS4s.

You keep posting high performance specs for a storage solution that is 2-3x (200-300% more) the price, 3x (300%) the power draw and heat, 5-6x (500-600%) the size and only yields a 25% increase in performance for what is needed. That's the point. It's like comparing a Ferrari to a Mustang. For the space and cost of an NVMe solution Sony could put multiple chips with their own bus and it would perform better in every metric including price. The NVMe controller alone pulls more power than the entire UFS 3.0 solution. Further performance of NVMe as a whole degrades as much as 25% as the solution warms up over the course of 30 mins due to its excessive power draw. After 30 mins, performance of most NVMes becomes erratic or degraded.

Avatar image for michaelmikado
michaelmikado

406

Forum Posts

0

Wiki Points

0

Followers

Reviews: 0

User Lists: 5

#227 michaelmikado
Member since 2019 • 406 Posts

@rzxv04 said:

Thanks, you two.

So I guess it'll depend on better price and mass availability.

NVMEs seem to be dropping fast and if the trend continues may have a price advantage from $ 60 256GB UFS 3.0 but NVME has worse power consumption and space (by how much?).

That PM961 seems expensive.

Would Sony have to pay extra for paying for UFS 3.0 controllers?

Not sure how practical this comparison is 7 Pro vs XS Max

https://www.youtube.com/watch?v=Rs2bmYhW738

I'm surprised how many times the android was faster and from what I know it's bottom tier in terms of dev optimization for majority of apps. Maybe if Apple adopts UFS 3.0 we'll be able to see it really shine because devs will probably optimize more for apple apps.

@rzxv04 said:

1. Makes sense.

2. Said "any ssd available for PC".

Would it be possible for them to currently have the highest bandwidth for a current mass market ssd of but somehow slower in other metrics? Something like wide but slower?

UFS 3.0 is exactly the solution you are talking about.

Here is a document from a couple years ago about the bus.

https://www.flashmemorysummit.com/English/Collaterals/Proceedings/2017/20170810_FL32_03_Tsai.pdf

It takes the same approach as HBM2 where instead of small high power, high heat busses, it employs very very width low power and small buses. UFS 3.0 may or may not come to apple but it will come to the billions of android devices, Car systems, Chromebooks, and low-end device computer devices. Its cost vs performance is the key to its widespread availability.

This is the quote from the document I posted and exactly what I keep trying to emphasize:

In device controller design, the balance between performance, power and cost is critical. A total control of the design will offer more flexibility to optimize the solution

Avatar image for rzxv04
rzxv04

2578

Forum Posts

0

Wiki Points

0

Followers

Reviews: 0

User Lists: 10

#228 rzxv04
Member since 2018 • 2578 Posts

@michaelmikado said:
@rzxv04 said:

Thanks, you two.

So I guess it'll depend on better price and mass availability.

NVMEs seem to be dropping fast and if the trend continues may have a price advantage from $ 60 256GB UFS 3.0 but NVME has worse power consumption and space (by how much?).

That PM961 seems expensive.

Would Sony have to pay extra for paying for UFS 3.0 controllers?

Not sure how practical this comparison is 7 Pro vs XS Max

https://www.youtube.com/watch?v=Rs2bmYhW738

I'm surprised how many times the android was faster and from what I know it's bottom tier in terms of dev optimization for majority of apps. Maybe if Apple adopts UFS 3.0 we'll be able to see it really shine because devs will probably optimize more for apple apps.

@rzxv04 said:

1. Makes sense.

2. Said "any ssd available for PC".

Would it be possible for them to currently have the highest bandwidth for a current mass market ssd of but somehow slower in other metrics? Something like wide but slower?

UFS 3.0 is exactly the solution you are talking about.

Here is a document from a couple years ago about the bus.

https://www.flashmemorysummit.com/English/Collaterals/Proceedings/2017/20170810_FL32_03_Tsai.pdf

It takes the same approach as HBM2 where instead of small high power, high heat busses, it employs very very width low power and small buses. UFS 3.0 may or may not come to apple but it will come to the billions of android devices, Car systems, Chromebooks, and low-end device computer devices. Its cost vs performance is the key to its widespread availability.

This is the quote from the document I posted and exactly what I keep trying to emphasize:

In device controller design, the balance between performance, power and cost is critical. A total control of the design will offer more flexibility to optimize the solution

Thanks for being patient with me.

Does this mean SSD is doomed?

Avatar image for michaelmikado
michaelmikado

406

Forum Posts

0

Wiki Points

0

Followers

Reviews: 0

User Lists: 5

#229 michaelmikado
Member since 2019 • 406 Posts

@rzxv04:

Not at all, the future is pretty much all 3D NAND for the foreseeable future. NVMe is still a superior yet immature, costly, and power hungry solution. It has a place in mid to high end devices which prioritize performance and it will improve substantially in the future.

Think of it like this.

Low end, low power devices = UFS

High power, expensive devices = NVMe

Thinking about it this way the PS5 will cost hundreds of dollars less than a high end PC or iPhone. It will likely have the lower tiers of next gen hardware which will still outperform much of the mid to top tier current gen hardware. None of this is a bad thing and it allows them to reduce the costs of the device later in its life rather than relying on highly niche expensive parts. Both solutions have a place. Plus if you think about it, it paves the way for something like a portable console similar to the switch in later revisions.

Avatar image for rzxv04
rzxv04

2578

Forum Posts

0

Wiki Points

0

Followers

Reviews: 0

User Lists: 10

#230 rzxv04
Member since 2018 • 2578 Posts

@michaelmikado said:

@rzxv04:

Not at all, the future is pretty much all 3D NAND for the foreseeable future. NVMe is still a superior yet immature, costly, and power hungry solution. It has a place in mid to high end devices which prioritize performance and it will improve substantially in the future.

Think of it like this.

Low end, low power devices = UFS

High power, expensive devices = NVMe

Thinking about it this way the PS5 will cost hundreds of dollars less than a high end PC or iPhone. It will likely have the lower tiers of next gen hardware which will still outperform much of the mid to top tier current gen hardware. None of this is a bad thing and it allows them to reduce the costs of the device later in its life rather than relying on highly niche expensive parts. Both solutions have a place. Plus if you think about it, it paves the way for something like a portable console similar to the switch in later revisions.

Forgive me but didn't you meantion that UFS 3.0 already performs close to nvm?

@michaelmikado said:

@rzxv04:

Not at all, the future is pretty much all 3D NAND for the foreseeable future. NVMe is still a superior yet immature, costly, and power hungry solution. It has a place in mid to high end devices which prioritize performance and it will improve substantially in the future.

Think of it like this.

Low end, low power devices = UFS

High power, expensive devices = NVMe

Thinking about it this way the PS5 will cost hundreds of dollars less than a high end PC or iPhone. It will likely have the lower tiers of next gen hardware which will still outperform much of the mid to top tier current gen hardware. None of this is a bad thing and it allows them to reduce the costs of the device later in its life rather than relying on highly niche expensive parts. Both solutions have a place. Plus if you think about it, it paves the way for something like a portable console similar to the switch in later revisions.

Btw what was that chart by ron that shows 220mbps while pc nvme does 1400 minimum sustained?

As for price isn't 256gb at $ 60 the more expensive compared to a lower end 1TB x4 nvme at $ 100?

If UFS really has close to nvme performance (3.2 nvme vs 2.9 ufs 3), and much cheaper, shouldn't the entire industry, including PC move towards it asap or do you think UFS controller makers would charge premiums?

It's also interesting that you did bring up that "raw bandwidth higher than any SSD available for PC" qualifier that makes UFS 3.0 interesting in that context but what do you think of a theoretical, very fast cache, low amount of storage system. Might require too much involvement and hassle for programmers?

Do you think Sony is somewhat being "coy/smart" that they meant they have a PCIE 4.0 in their system with a soon to be released but soon to be common mid-high end NVME that takes advantage of PCIE 4.0 in their dev kits for a few months now? Technically faster than any mass market ssds today.. but maybe common enough in late 2019 or entrance of 2020?

Avatar image for michaelmikado
michaelmikado

406

Forum Posts

0

Wiki Points

0

Followers

Reviews: 0

User Lists: 5

#231 michaelmikado
Member since 2019 • 406 Posts

@rzxv04 said:
@michaelmikado said:

@rzxv04:

Not at all, the future is pretty much all 3D NAND for the foreseeable future. NVMe is still a superior yet immature, costly, and power hungry solution. It has a place in mid to high end devices which prioritize performance and it will improve substantially in the future.

Think of it like this.

Low end, low power devices = UFS

High power, expensive devices = NVMe

Thinking about it this way the PS5 will cost hundreds of dollars less than a high end PC or iPhone. It will likely have the lower tiers of next gen hardware which will still outperform much of the mid to top tier current gen hardware. None of this is a bad thing and it allows them to reduce the costs of the device later in its life rather than relying on highly niche expensive parts. Both solutions have a place. Plus if you think about it, it paves the way for something like a portable console similar to the switch in later revisions.

Forgive me but didn't you meantion that UFS 3.0 already performs close to nvm?

@michaelmikado said:

@rzxv04:

Not at all, the future is pretty much all 3D NAND for the foreseeable future. NVMe is still a superior yet immature, costly, and power hungry solution. It has a place in mid to high end devices which prioritize performance and it will improve substantially in the future.

Think of it like this.

Low end, low power devices = UFS

High power, expensive devices = NVMe

Thinking about it this way the PS5 will cost hundreds of dollars less than a high end PC or iPhone. It will likely have the lower tiers of next gen hardware which will still outperform much of the mid to top tier current gen hardware. None of this is a bad thing and it allows them to reduce the costs of the device later in its life rather than relying on highly niche expensive parts. Both solutions have a place. Plus if you think about it, it paves the way for something like a portable console similar to the switch in later revisions.

Btw what was that chart by ron that shows 220mbps while pc nvme does 1400 minimum sustained?

As for price isn't 256gb at $ 60 the more expensive compared to a lower end 1TB x4 nvme at $ 100?

If UFS really has close to nvme performance (3.2 nvme vs 2.9 ufs 3), and much cheaper, shouldn't the entire industry, including PC move towards it asap or do you think UFS controller makers would charge premiums?

It's also interesting that you did bring up that "raw bandwidth higher than any SSD available for PC" qualifier that makes UFS 3.0 interesting in that context but what do you think of a theoretical, very fast cache, low amount of storage system. Might require too much involvement and hassle for programmers?

Do you think Sony is somewhat being "coy/smart" that they meant they have a PCIE 4.0 in their system with a soon to be released but soon to be common mid-high end NVME that takes advantage of PCIE 4.0 in their dev kits for a few months now? Technically faster than any mass market ssds today.. but maybe common enough in late 2019 or entrance of 2020?

With anything in there are always trade-offs. UFS 3.0 has low write speeds and random reads when compared to NVMe. This makes NVMe far far far better suited in a PC environment where you do multiple things at high performance. As opposed to a cellphone, Chromebook, car driving system, where you do 1 task which is to act as a fast cache. That's why UFS 3.0 isn't suited to general PC use for mid and high-end performance. It literally does one thing very fast for very cheap and its literally to act as fast cache. The one and only benefit Sony touted. Realistically there is no reason to have very high write speeds or random reads in something that will predominantly be a linear experience only needing a cache. Further high write speeds do not make sense when download speeds and installs from blu-ray would be very low anyway. There's simple no point in paying for all the draw additional drawbacks to get benefits that you don't need or use. Remember UFS 3.0 is superior as a cheap, low power cache, not a full on PC HDD. There's no premiums for UFS 3.0 controllers because the do not have nearly as much logic built into them like NVMe.

As for the price, I was just using it as an example. Realistically a 512GB drive would likely cost something like $30-$40 for Sony. Also it isn't looking at it as value, rather it would be raw cost. A 512GB drive at $40 vs a 1TB drive at $60 is still $20 more in costs which still doesn't account for added costs of the controller, the heat solutions, and the reduction in power to the GPU, memory, or GPU. For that same $20 cost plus the controller and other costs they could get a better processor, more RAM, invest more in the GPU etc.

To kind of put this in perspective, the cost to put NVMe doesn't add much to the actually applications Sony is looking to use it as which is basically a graduated cache. As for whether they would benefit from small amounts, I originally they would have a eMMC drive between 32 and 128GB for this purpose, but since then Samsung who makes much of the PS4/5 RAM has suspiciously spun up their UFS 3.0 mass production and their minimum size I believe is 128/256GB. They still could have a custom solution which is even faster than normal and in small capacities. We just have to wait for them to officially make that claim.

Avatar image for rzxv04
rzxv04

2578

Forum Posts

0

Wiki Points

0

Followers

Reviews: 0

User Lists: 10

#232 rzxv04
Member since 2018 • 2578 Posts

@michaelmikado said:
@rzxv04 said:
@michaelmikado said:

@rzxv04:

Not at all, the future is pretty much all 3D NAND for the foreseeable future. NVMe is still a superior yet immature, costly, and power hungry solution. It has a place in mid to high end devices which prioritize performance and it will improve substantially in the future.

Think of it like this.

Low end, low power devices = UFS

High power, expensive devices = NVMe

Thinking about it this way the PS5 will cost hundreds of dollars less than a high end PC or iPhone. It will likely have the lower tiers of next gen hardware which will still outperform much of the mid to top tier current gen hardware. None of this is a bad thing and it allows them to reduce the costs of the device later in its life rather than relying on highly niche expensive parts. Both solutions have a place. Plus if you think about it, it paves the way for something like a portable console similar to the switch in later revisions.

Forgive me but didn't you meantion that UFS 3.0 already performs close to nvm?

@michaelmikado said:

@rzxv04:

Not at all, the future is pretty much all 3D NAND for the foreseeable future. NVMe is still a superior yet immature, costly, and power hungry solution. It has a place in mid to high end devices which prioritize performance and it will improve substantially in the future.

Think of it like this.

Low end, low power devices = UFS

High power, expensive devices = NVMe

Thinking about it this way the PS5 will cost hundreds of dollars less than a high end PC or iPhone. It will likely have the lower tiers of next gen hardware which will still outperform much of the mid to top tier current gen hardware. None of this is a bad thing and it allows them to reduce the costs of the device later in its life rather than relying on highly niche expensive parts. Both solutions have a place. Plus if you think about it, it paves the way for something like a portable console similar to the switch in later revisions.

Btw what was that chart by ron that shows 220mbps while pc nvme does 1400 minimum sustained?

As for price isn't 256gb at $ 60 the more expensive compared to a lower end 1TB x4 nvme at $ 100?

If UFS really has close to nvme performance (3.2 nvme vs 2.9 ufs 3), and much cheaper, shouldn't the entire industry, including PC move towards it asap or do you think UFS controller makers would charge premiums?

It's also interesting that you did bring up that "raw bandwidth higher than any SSD available for PC" qualifier that makes UFS 3.0 interesting in that context but what do you think of a theoretical, very fast cache, low amount of storage system. Might require too much involvement and hassle for programmers?

Do you think Sony is somewhat being "coy/smart" that they meant they have a PCIE 4.0 in their system with a soon to be released but soon to be common mid-high end NVME that takes advantage of PCIE 4.0 in their dev kits for a few months now? Technically faster than any mass market ssds today.. but maybe common enough in late 2019 or entrance of 2020?

With anything in there are always trade-offs. UFS 3.0 has low write speeds and random reads when compared to NVMe. This makes NVMe far far far better suited in a PC environment where you do multiple things at high performance. As opposed to a cellphone, Chromebook, car driving system, where you do 1 task which is to act as a fast cache. That's why UFS 3.0 isn't suited to general PC use for mid and high-end performance. It literally does one thing very fast for very cheap and its literally to act as fast cache. The one and only benefit Sony touted. Realistically there is no reason to have very high write speeds or random reads in something that will predominantly be a linear experience only needing a cache. Further high write speeds do not make sense when download speeds and installs from blu-ray would be very low anyway. There's simple no point in paying for all the draw additional drawbacks to get benefits that you don't need or use. Remember UFS 3.0 is superior as a cheap, low power cache, not a full on PC HDD. There's no premiums for UFS 3.0 controllers because the do not have nearly as much logic built into them like NVMe.

As for the price, I was just using it as an example. Realistically a 512GB drive would likely cost something like $30-$40 for Sony. Also it isn't looking at it as value, rather it would be raw cost. A 512GB drive at $40 vs a 1TB drive at $60 is still $20 more in costs which still doesn't account for added costs of the controller, the heat solutions, and the reduction in power to the GPU, memory, or GPU. For that same $20 cost plus the controller and other costs they could get a better processor, more RAM, invest more in the GPU etc.

To kind of put this in perspective, the cost to put NVMe doesn't add much to the actually applications Sony is looking to use it as which is basically a graduated cache. As for whether they would benefit from small amounts, I originally they would have a eMMC drive between 32 and 128GB for this purpose, but since then Samsung who makes much of the PS4/5 RAM has suspiciously spun up their UFS 3.0 mass production and their minimum size I believe is 128/256GB. They still could have a custom solution which is even faster than normal and in small capacities. We just have to wait for them to officially make that claim.

Do you think there's a good chance of heavily investing into video/streamed editing built in the PS5 to make use of NVME being faster all around?

Would you happen to have figures of how much power consumption advantage UFS 3.0 has over nvme?

Do you think that nvme's heat isn't much of an issue? Would slapping small heatsinks be good enough? I don't see those slapped onto PCs yet.

Is this exaggerated?:

I can’t stress this enough – if you’re using the M.2 NVMe SSD as a boot drive and for gaming/daily use – then don’t worry, it’s really difficult to make it thermal throttle unless it’s copying files in and out continuously. It’s reallydifficult to create such a scenario.

https://nasilemaktech.com/nvme-ssd-thermal-throttling-preventions/

Just being near heatsinks seem to cool it off?

By the way where does the controller reside? Is that already part of the cheap x4 1TB $ 100 stick or would that be a separate chip on the motherboard?

In what scenario do you think the PS5 wouldn't use UFS 3.0?

One rumor was talking about Phison? How is that compared to UFS?

Avatar image for ronvalencia
ronvalencia

29612

Forum Posts

0

Wiki Points

0

Followers

Reviews: 1

User Lists: 0

#233  Edited By ronvalencia
Member since 2008 • 29612 Posts

@michaelmikado said:
@ronvalencia said:
@michaelmikado said:
@rzxv04 said:

Thanks.

I can't seem to understand this guy's presentation. What exactly does he mean that UFS 3.0 isn't as fast?

https://www.youtube.com/watch?v=TaKC1qIlo9Y&t=572s

NVME can seem to come down in price through lots of sales the past few months about $ 100 / TB:

https://slickdeals.net/deals/ssd/

https://www.microcenter.com/product/600422/1TB_SSD_3D_NAND_M2_2280_PCIe_NVMe_30_x4_Internal_Solid_State_Drive

I imagine bulk purchases by Sony and MS would be much cheaper.

How inferior is NVME x2 compared to UFS 3.0?

He is referencing sequential reads vs random reads. NVMe excels at random reads. UFS excels at sequential. Even at the lowest real world estimates for UFS 3.0 it out paces the highest theoretical performance for NVMe x2 at 1.6-1.8GBps in sequential reads but UFS suffers in random reads. Random reads really benefit exactly from what it sounds like. Unpredictable activity such as launching apps by the end user while sequential reads would be something like 4K video where the information is kept in blocks located close together on the storage media. Again, in a console you do not have a bunch of apps opening. The console has one primary function so it will be relatively easy to manage memory. Even in open world games there is only a finite amount you need to swap into memory at high speed and any load times that would occur would be so minimal that a developer could mask them easily even for something for fast travel. Basically it would enable load times in the single digits if even that for far cheaper and less cost all around and more wattage can be allocated to the GPU or RAM which is vitally important. You wouldn't want an SSD to suck up a significant amount of cost, power, or space at the expense of everything else just to get a second or two less of already low load times. That's what consoles are trade offs because there comes a power where solutions are "good enough" and balance out for the best for the consoles.

NVMe excels in both random and sequential

https://www.tweaktown.com/reviews/8176/samsung-pm961-1tb-2-nvme-pcie-ssd-review/index4.html

Sequential read reached ~3.1 GB/sbenchmarked (not theoretical) for Samsung PM961 nvme.

Highest theoretical performance is useless. Ryzen's results are slightly less than Intel's.

Sony has used laptop 2.5 inch HDDs in PS4s.

You keep posting high performance specs for a storage solution that is 2-3x (200-300% more) the price, 3x (300%) the power draw and heat, 5-6x (500-600%) the size and only yields a 25% increase in performance for what is needed. That's the point. It's like comparing a Ferrari to a Mustang. For the space and cost of an NVMe solution Sony could put multiple chips with their own bus and it would perform better in every metric including price. The NVMe controller alone pulls more power than the entire UFS 3.0 solution. Further performance of NVMe as a whole degrades as much as 25% as the solution warms up over the course of 30 mins due to its excessive power draw. After 30 mins, performance of most NVMes becomes erratic or degraded.

Sony has tolerated laptop 2.5 inch HDDs. NVMe M2 is used Surface Pro 4 tablets. My 2-in-1 tablet/laptop has NVMe M2.

https://pcpartpicker.com/product/xpYLrH/crucial-p1-500gb-m2-2280-solid-state-drive-ct500p1ssd8

Crucial P1 with 500 GB M.2-2280 for $59.99 retail will be cheaper at wholesale or direct from manufacture.

Surface Pro 5 reduced power consumption by removing M2 slot infrastructure and it's still NVMe.

M.2 scales up to four PCI-E version 3.0 lanes. Next version is based on PCI-E version 4.0 which comes included with Zen v2, but PS5's 18X over PS4 Pro's results shows lower cost Crucial P1 like storage chips, hence the bottleneck is with storage chips. The bottleneck is cost targets for PS5 despite Zen v2's I/O improvements.

PS5 with multiple 8 core Zen v2 CPU cores would have heavy multi-threaded storage read operations with internet client service stack, hence it's like a PC. PS4's 3 GB OS reserve memory is worst than Windows 10 with 2GB RAM tablet.

----------

https://www.techpowerup.com/255826/intel-sapphire-rapids-brings-pcie-gen-5-and-ddr5-to-the-data-center

Intel's PCI-E version 5.0 road map for year 2021.

Avatar image for michaelmikado
michaelmikado

406

Forum Posts

0

Wiki Points

0

Followers

Reviews: 0

User Lists: 5

#234 michaelmikado
Member since 2019 • 406 Posts

@rzxv04 said:
@michaelmikado said:
@rzxv04 said:
@michaelmikado said:

@rzxv04:

Not at all, the future is pretty much all 3D NAND for the foreseeable future. NVMe is still a superior yet immature, costly, and power hungry solution. It has a place in mid to high end devices which prioritize performance and it will improve substantially in the future.

Think of it like this.

Low end, low power devices = UFS

High power, expensive devices = NVMe

Thinking about it this way the PS5 will cost hundreds of dollars less than a high end PC or iPhone. It will likely have the lower tiers of next gen hardware which will still outperform much of the mid to top tier current gen hardware. None of this is a bad thing and it allows them to reduce the costs of the device later in its life rather than relying on highly niche expensive parts. Both solutions have a place. Plus if you think about it, it paves the way for something like a portable console similar to the switch in later revisions.

Forgive me but didn't you meantion that UFS 3.0 already performs close to nvm?

@michaelmikado said:

@rzxv04:

Not at all, the future is pretty much all 3D NAND for the foreseeable future. NVMe is still a superior yet immature, costly, and power hungry solution. It has a place in mid to high end devices which prioritize performance and it will improve substantially in the future.

Think of it like this.

Low end, low power devices = UFS

High power, expensive devices = NVMe

Thinking about it this way the PS5 will cost hundreds of dollars less than a high end PC or iPhone. It will likely have the lower tiers of next gen hardware which will still outperform much of the mid to top tier current gen hardware. None of this is a bad thing and it allows them to reduce the costs of the device later in its life rather than relying on highly niche expensive parts. Both solutions have a place. Plus if you think about it, it paves the way for something like a portable console similar to the switch in later revisions.

Btw what was that chart by ron that shows 220mbps while pc nvme does 1400 minimum sustained?

As for price isn't 256gb at $ 60 the more expensive compared to a lower end 1TB x4 nvme at $ 100?

If UFS really has close to nvme performance (3.2 nvme vs 2.9 ufs 3), and much cheaper, shouldn't the entire industry, including PC move towards it asap or do you think UFS controller makers would charge premiums?

It's also interesting that you did bring up that "raw bandwidth higher than any SSD available for PC" qualifier that makes UFS 3.0 interesting in that context but what do you think of a theoretical, very fast cache, low amount of storage system. Might require too much involvement and hassle for programmers?

Do you think Sony is somewhat being "coy/smart" that they meant they have a PCIE 4.0 in their system with a soon to be released but soon to be common mid-high end NVME that takes advantage of PCIE 4.0 in their dev kits for a few months now? Technically faster than any mass market ssds today.. but maybe common enough in late 2019 or entrance of 2020?

With anything in there are always trade-offs. UFS 3.0 has low write speeds and random reads when compared to NVMe. This makes NVMe far far far better suited in a PC environment where you do multiple things at high performance. As opposed to a cellphone, Chromebook, car driving system, where you do 1 task which is to act as a fast cache. That's why UFS 3.0 isn't suited to general PC use for mid and high-end performance. It literally does one thing very fast for very cheap and its literally to act as fast cache. The one and only benefit Sony touted. Realistically there is no reason to have very high write speeds or random reads in something that will predominantly be a linear experience only needing a cache. Further high write speeds do not make sense when download speeds and installs from blu-ray would be very low anyway. There's simple no point in paying for all the draw additional drawbacks to get benefits that you don't need or use. Remember UFS 3.0 is superior as a cheap, low power cache, not a full on PC HDD. There's no premiums for UFS 3.0 controllers because the do not have nearly as much logic built into them like NVMe.

As for the price, I was just using it as an example. Realistically a 512GB drive would likely cost something like $30-$40 for Sony. Also it isn't looking at it as value, rather it would be raw cost. A 512GB drive at $40 vs a 1TB drive at $60 is still $20 more in costs which still doesn't account for added costs of the controller, the heat solutions, and the reduction in power to the GPU, memory, or GPU. For that same $20 cost plus the controller and other costs they could get a better processor, more RAM, invest more in the GPU etc.

To kind of put this in perspective, the cost to put NVMe doesn't add much to the actually applications Sony is looking to use it as which is basically a graduated cache. As for whether they would benefit from small amounts, I originally they would have a eMMC drive between 32 and 128GB for this purpose, but since then Samsung who makes much of the PS4/5 RAM has suspiciously spun up their UFS 3.0 mass production and their minimum size I believe is 128/256GB. They still could have a custom solution which is even faster than normal and in small capacities. We just have to wait for them to officially make that claim.

Do you think there's a good chance of heavily investing into video/streamed editing built in the PS5 to make use of NVME being faster all around?

Would you happen to have figures of how much power consumption advantage UFS 3.0 has over nvme?

Do you think that nvme's heat isn't much of an issue? Would slapping small heatsinks be good enough? I don't see those slapped onto PCs yet.

Is this exaggerated?:

I can’t stress this enough – if you’re using the M.2 NVMe SSD as a boot drive and for gaming/daily use – then don’t worry, it’s really difficult to make it thermal throttle unless it’s copying files in and out continuously. It’s reallydifficult to create such a scenario.

https://nasilemaktech.com/nvme-ssd-thermal-throttling-preventions/

Just being near heatsinks seem to cool it off?

By the way where does the controller reside? Is that already part of the cheap x4 1TB $ 100 stick or would that be a separate chip on the motherboard?

In what scenario do you think the PS5 wouldn't use UFS 3.0?

One rumor was talking about Phison? How is that compared to UFS?

Any video would likely work like how the PS4 does. Actively stored in memory and transferred to disk when needed. It may be part of the needed (rumored) 4GB of RAM dedicated to the OS. There's no sense in writing any streaming content to disk especially with that much RAM.

UFS has a max of 1.68 Watts power consumption. By contrast Optane gets up to 17+ W during writes and around 10W reads. Most NVMe should be around 10W and below max though. But still 5-10x the power consumption. I would say at least 3x as much minimum.

The quote about throttling isn't exaggerated because of the way Operating Systems and Games work where they load everything into RAM in one shot and then the Disk will rest or go idle for a bit because everything is stored in memory. The reason games and OSs are made that way is because they cannot rely on fast drives to operate. Sony is planning the exact opposite as they want to use it as an active cache. Thus the specific scenario which they are saying it would actually be a problem. it’s really difficult to make it thermal throttle unless it’s copying files in and out continuously.Is the exact use Sony has planned for the PS5 drive. Its not just about load times being shorter they want to eliminate them completely by letting devs has access to continuous and reliably fast storage. You can't do that if your storage throttles itself after 30 minutes of use.

I would say the problem is that you have to cool it off to begin with. NVMe Max operating temps are 185F by contrast UFS can get to 225F. UFS could basically sit on the processor and still survive and not throttle. I don't think any console manufacturer would want to risk a red ring of death scenario again.

The controller for both is typically on the storage itself. It's important to remember that the controller is really the only thing that separates UFS 3.0 from NVMe, as its just the interface. Both use Flash NAND and the interfaces just prioritize two different methods to access it. That said UFS 3.0 also has an embedded solution which would allow you to drop multiple block of Flash NAND directly on the board without working about PCI-E channels and busses. For both you would need the actually connectors and busses and m.2 is small but not smaller than UFS because one is obviously designed for mobile and is more limited.

I have no idea if they actually would use UFS or not, but if I was making a console that's the one I would chose for all the listed reasons. In order for it to be viable it would need to 1) outperform UFS 2) be cheaper than UFS 3) run cooler and less power than UFS. Which by the time you get to this point on costs with NVMe you are failing in both performance and power. As an example the link to the crucial NVMe storage in the other link has a max of 2GBps. When you reduce costs or power on NVMe you lose performance making it worse than UFS on sequential reads which is what you actually need. If you increase performance you pay too much and it runs to hot and ends up getting throttled back down anyway.

I could see them using NVMe if they just weren't able to workout anything better on the business end.

Yeah I heard about the Phison controller at computex or something. Don't get me wrong they make excellent controllers and theirs will be something to see, but at the end of the day its going to be high-end in price and power so it wouldn't work for a console. The main consideration is balancing the solutions.

Here is what I would do if I were Sony:

UFS 3.0 is merely a standard. And the standard was developed to enhance performance in low power scenarios, hence the very low wattage. Sony could easily work with a company to use the same chips but a modified UFS 3.0 controller which doubles power consumption placing in the realm of a low power NVMe and easily exceed 4/5GBps. Alternatively, the controllers are cheap because UFS mostly scales linearly. NVMe has more advanced logic built into the controller hence why they perform better. UFS by contrast is pretty dumb and requires slightly more cpu resources to management. Sony is rumored to have developed an advanced memory logic system to automate this process (Also why NVMe would be redundant). Sony, in theory could drop multiple UFS 3.0 blocks on the board instead. A 128GB block would likely be around $10-$15. They could drop 4 of these, have 512GB of storage with each 128GB block having its own 2.9GBps of bandwidth and let its memory management system sort out the caching. This is all hypothetical of course so don't assume anything like that would happen, but a solution like that would technically be leagues better than anything on PC at the moment.

Avatar image for fiziwee
fiziwee

8

Forum Posts

0

Wiki Points

0

Followers

Reviews: 0

User Lists: 5

#235  Edited By fiziwee
Member since 2019 • 8 Posts

Xbox doesn't have Loading screens right?

Avatar image for rzxv04
rzxv04

2578

Forum Posts

0

Wiki Points

0

Followers

Reviews: 0

User Lists: 10

#236 rzxv04
Member since 2018 • 2578 Posts

@michaelmikado said:
@rzxv04 said:
@michaelmikado said:
@rzxv04 said:
@michaelmikado said:

@rzxv04:

Not at all, the future is pretty much all 3D NAND for the foreseeable future. NVMe is still a superior yet immature, costly, and power hungry solution. It has a place in mid to high end devices which prioritize performance and it will improve substantially in the future.

Think of it like this.

Low end, low power devices = UFS

High power, expensive devices = NVMe

Thinking about it this way the PS5 will cost hundreds of dollars less than a high end PC or iPhone. It will likely have the lower tiers of next gen hardware which will still outperform much of the mid to top tier current gen hardware. None of this is a bad thing and it allows them to reduce the costs of the device later in its life rather than relying on highly niche expensive parts. Both solutions have a place. Plus if you think about it, it paves the way for something like a portable console similar to the switch in later revisions.

Forgive me but didn't you meantion that UFS 3.0 already performs close to nvm?

@michaelmikado said:

@rzxv04:

Not at all, the future is pretty much all 3D NAND for the foreseeable future. NVMe is still a superior yet immature, costly, and power hungry solution. It has a place in mid to high end devices which prioritize performance and it will improve substantially in the future.

Think of it like this.

Low end, low power devices = UFS

High power, expensive devices = NVMe

Thinking about it this way the PS5 will cost hundreds of dollars less than a high end PC or iPhone. It will likely have the lower tiers of next gen hardware which will still outperform much of the mid to top tier current gen hardware. None of this is a bad thing and it allows them to reduce the costs of the device later in its life rather than relying on highly niche expensive parts. Both solutions have a place. Plus if you think about it, it paves the way for something like a portable console similar to the switch in later revisions.

Btw what was that chart by ron that shows 220mbps while pc nvme does 1400 minimum sustained?

As for price isn't 256gb at $ 60 the more expensive compared to a lower end 1TB x4 nvme at $ 100?

If UFS really has close to nvme performance (3.2 nvme vs 2.9 ufs 3), and much cheaper, shouldn't the entire industry, including PC move towards it asap or do you think UFS controller makers would charge premiums?

It's also interesting that you did bring up that "raw bandwidth higher than any SSD available for PC" qualifier that makes UFS 3.0 interesting in that context but what do you think of a theoretical, very fast cache, low amount of storage system. Might require too much involvement and hassle for programmers?

Do you think Sony is somewhat being "coy/smart" that they meant they have a PCIE 4.0 in their system with a soon to be released but soon to be common mid-high end NVME that takes advantage of PCIE 4.0 in their dev kits for a few months now? Technically faster than any mass market ssds today.. but maybe common enough in late 2019 or entrance of 2020?

With anything in there are always trade-offs. UFS 3.0 has low write speeds and random reads when compared to NVMe. This makes NVMe far far far better suited in a PC environment where you do multiple things at high performance. As opposed to a cellphone, Chromebook, car driving system, where you do 1 task which is to act as a fast cache. That's why UFS 3.0 isn't suited to general PC use for mid and high-end performance. It literally does one thing very fast for very cheap and its literally to act as fast cache. The one and only benefit Sony touted. Realistically there is no reason to have very high write speeds or random reads in something that will predominantly be a linear experience only needing a cache. Further high write speeds do not make sense when download speeds and installs from blu-ray would be very low anyway. There's simple no point in paying for all the draw additional drawbacks to get benefits that you don't need or use. Remember UFS 3.0 is superior as a cheap, low power cache, not a full on PC HDD. There's no premiums for UFS 3.0 controllers because the do not have nearly as much logic built into them like NVMe.

As for the price, I was just using it as an example. Realistically a 512GB drive would likely cost something like $30-$40 for Sony. Also it isn't looking at it as value, rather it would be raw cost. A 512GB drive at $40 vs a 1TB drive at $60 is still $20 more in costs which still doesn't account for added costs of the controller, the heat solutions, and the reduction in power to the GPU, memory, or GPU. For that same $20 cost plus the controller and other costs they could get a better processor, more RAM, invest more in the GPU etc.

To kind of put this in perspective, the cost to put NVMe doesn't add much to the actually applications Sony is looking to use it as which is basically a graduated cache. As for whether they would benefit from small amounts, I originally they would have a eMMC drive between 32 and 128GB for this purpose, but since then Samsung who makes much of the PS4/5 RAM has suspiciously spun up their UFS 3.0 mass production and their minimum size I believe is 128/256GB. They still could have a custom solution which is even faster than normal and in small capacities. We just have to wait for them to officially make that claim.

Do you think there's a good chance of heavily investing into video/streamed editing built in the PS5 to make use of NVME being faster all around?

Would you happen to have figures of how much power consumption advantage UFS 3.0 has over nvme?

Do you think that nvme's heat isn't much of an issue? Would slapping small heatsinks be good enough? I don't see those slapped onto PCs yet.

Is this exaggerated?:

I can’t stress this enough – if you’re using the M.2 NVMe SSD as a boot drive and for gaming/daily use – then don’t worry, it’s really difficult to make it thermal throttle unless it’s copying files in and out continuously. It’s reallydifficult to create such a scenario.

https://nasilemaktech.com/nvme-ssd-thermal-throttling-preventions/

Just being near heatsinks seem to cool it off?

By the way where does the controller reside? Is that already part of the cheap x4 1TB $ 100 stick or would that be a separate chip on the motherboard?

In what scenario do you think the PS5 wouldn't use UFS 3.0?

One rumor was talking about Phison? How is that compared to UFS?

Any video would likely work like how the PS4 does. Actively stored in memory and transferred to disk when needed. It may be part of the needed (rumored) 4GB of RAM dedicated to the OS. There's no sense in writing any streaming content to disk especially with that much RAM.

UFS has a max of 1.68 Watts power consumption. By contrast Optane gets up to 17+ W during writes and around 10W reads. Most NVMe should be around 10W and below max though. But still 5-10x the power consumption. I would say at least 3x as much minimum.

The quote about throttling isn't exaggerated because of the way Operating Systems and Games work where they load everything into RAM in one shot and then the Disk will rest or go idle for a bit because everything is stored in memory. The reason games and OSs are made that way is because they cannot rely on fast drives to operate. Sony is planning the exact opposite as they want to use it as an active cache. Thus the specific scenario which they are saying it would actually be a problem. it’s really difficult to make it thermal throttle unless it’s copying files in and out continuously.Is the exact use Sony has planned for the PS5 drive. Its not just about load times being shorter they want to eliminate them completely by letting devs has access to continuous and reliably fast storage. You can't do that if your storage throttles itself after 30 minutes of use.

I would say the problem is that you have to cool it off to begin with. NVMe Max operating temps are 185F by contrast UFS can get to 225F. UFS could basically sit on the processor and still survive and not throttle. I don't think any console manufacturer would want to risk a red ring of death scenario again.

The controller for both is typically on the storage itself. It's important to remember that the controller is really the only thing that separates UFS 3.0 from NVMe, as its just the interface. Both use Flash NAND and the interfaces just prioritize two different methods to access it. That said UFS 3.0 also has an embedded solution which would allow you to drop multiple block of Flash NAND directly on the board without working about PCI-E channels and busses. For both you would need the actually connectors and busses and m.2 is small but not smaller than UFS because one is obviously designed for mobile and is more limited.

I have no idea if they actually would use UFS or not, but if I was making a console that's the one I would chose for all the listed reasons. In order for it to be viable it would need to 1) outperform UFS 2) be cheaper than UFS 3) run cooler and less power than UFS. Which by the time you get to this point on costs with NVMe you are failing in both performance and power. As an example the link to the crucial NVMe storage in the other link has a max of 2GBps. When you reduce costs or power on NVMe you lose performance making it worse than UFS on sequential reads which is what you actually need. If you increase performance you pay too much and it runs to hot and ends up getting throttled back down anyway.

I could see them using NVMe if they just weren't able to workout anything better on the business end.

Yeah I heard about the Phison controller at computex or something. Don't get me wrong they make excellent controllers and theirs will be something to see, but at the end of the day its going to be high-end in price and power so it wouldn't work for a console. The main consideration is balancing the solutions.

Here is what I would do if I were Sony:

UFS 3.0 is merely a standard. And the standard was developed to enhance performance in low power scenarios, hence the very low wattage. Sony could easily work with a company to use the same chips but a modified UFS 3.0 controller which doubles power consumption placing in the realm of a low power NVMe and easily exceed 4/5GBps. Alternatively, the controllers are cheap because UFS mostly scales linearly. NVMe has more advanced logic built into the controller hence why they perform better. UFS by contrast is pretty dumb and requires slightly more cpu resources to management. Sony is rumored to have developed an advanced memory logic system to automate this process (Also why NVMe would be redundant). Sony, in theory could drop multiple UFS 3.0 blocks on the board instead. A 128GB block would likely be around $10-$15. They could drop 4 of these, have 512GB of storage with each 128GB block having its own 2.9GBps of bandwidth and let its memory management system sort out the caching. This is all hypothetical of course so don't assume anything like that would happen, but a solution like that would technically be leagues better than anything on PC at the moment.

Keep posting. These are all interesting to me.

So UFS might need a slightly better CPU thus lessening some power for games but I guess these are hard to estimate.

I was thinking of high quality video editing built into the PS4 where I only heard that nvme's are taken advantage of copying huge files in and out for editing but I'm not sure how exactly that works. I just remember a youtuber mentioning that the PS4 Pro has horrible quality in their built in capture even compared to the Xbox One S. Was wondering if nvme could help with that but if that were actually an issue at all, UFS might do well enough.

From the looks of it, at least currently, since the controller that's supposed to be expensive is already part of the cheap package of even falling prices of NVME, seems barely an issue but we really need to see actual UFS pricing. Would you happen to have leads to find prices for these?

UFS does really seem like it makes better sense if Sony can instead use higher powered versions of it for the same heat/consumption budget of an NVME but maybe Sony's too late for this with the upcoming ps5.

Avatar image for michaelmikado
michaelmikado

406

Forum Posts

0

Wiki Points

0

Followers

Reviews: 0

User Lists: 5

#237  Edited By michaelmikado
Member since 2019 • 406 Posts

@rzxv04 said:

Keep posting. These are all interesting to me.

So UFS might need a slightly better CPU thus lessening some power for games but I guess these are hard to estimate.

I was thinking of high quality video editing built into the PS4 where I only heard that nvme's are taken advantage of copying huge files in and out for editing but I'm not sure how exactly that works. I just remember a youtuber mentioning that the PS4 Pro has horrible quality in their built in capture even compared to the Xbox One S. Was wondering if nvme could help with that but if that were actually an issue at all, UFS might do well enough.

From the looks of it, at least currently, since the controller that's supposed to be expensive is already part of the cheap package of even falling prices of NVME, seems barely an issue but we really need to see actual UFS pricing. Would you happen to have leads to find prices for these?

UFS does really seem like it makes better sense if Sony can instead use higher powered versions of it for the same heat/consumption budget of an NVME but maybe Sony's too late for this with the upcoming ps5.

Realistically UFS is going to use 0.00001% of a modern CPU, remember its something that runs flawlessly on low-end arm processors. UFS pricing will always be cheaper than NVMe pricing if the 3D Flash NAND is equivalent. Its because they use the same parts with the exception of the controller which always is more expensive on NVMe even on the lowest end models. Like I said the one above which was rated cheaper was only getting 2GBps, if you choose NVMe over UFS, the last thing you would want to do is go cheap and get an underperforming controller because it destroys any reason you had to get NVMe in the first place.

Interestingly enough these patents from Sony from 2015/2016 turned up on a couple of forums. http://www.freepatentsonline.com/y2017/0097897.html

The highlights are that it would use a small cache of SRAM instead of DRAM, it includes a hardware accelerator, and it focuses primarily on high read speeds rather than write speeds very similar to UFS. This goes back to what I said last time.

UFS 3.0 is merely a standard. And the standard was developed to enhance performance in low power scenarios, hence the very low wattage. Sony could easily work with a company to use the same chips but a modified UFS 3.0 controller which doubles power consumption placing in the realm of a low power NVMe and easily exceed 4/5GBps. Alternatively, the controllers are cheap because UFS mostly scales linearly. NVMe has more advanced logic built into the controller hence why they perform better. UFS by contrast is pretty dumb and requires slightly more cpu resources to management. Sony is rumored to have developed an advanced memory logic system to automate this process (Also why NVMe would be redundant). Sony, in theory could drop multiple UFS 3.0 blocks on the board instead. A 128GB block would likely be around $10-$15. They could drop 4 of these, have 512GB of storage with each 128GB block having its own 2.9GBps of bandwidth and let its memory management system sort out the caching.

My proposed off the cuff solution would have yielded a theoretical bandwidth of around 10GBps which is completely unheard of and not even in the near immediate future for PCs, but apparently Sony also worked out their math where much of their patent estimates are based around a 10GBps read speed as well so I guess I wasn't far off.

Also interestingly enough this also addresses the memory logic rumors with its secondary CPU and hardware accelerator built into the controller so its more likely Sony came to the same conclusions and decided to build an SSD controller specific to gaming applications rather than wasting resources of using something that was designed for a different application altogether.

Avatar image for rzxv04
rzxv04

2578

Forum Posts

0

Wiki Points

0

Followers

Reviews: 0

User Lists: 10

#238 rzxv04
Member since 2018 • 2578 Posts

@michaelmikado said:
@rzxv04 said:

Keep posting. These are all interesting to me.

So UFS might need a slightly better CPU thus lessening some power for games but I guess these are hard to estimate.

I was thinking of high quality video editing built into the PS4 where I only heard that nvme's are taken advantage of copying huge files in and out for editing but I'm not sure how exactly that works. I just remember a youtuber mentioning that the PS4 Pro has horrible quality in their built in capture even compared to the Xbox One S. Was wondering if nvme could help with that but if that were actually an issue at all, UFS might do well enough.

From the looks of it, at least currently, since the controller that's supposed to be expensive is already part of the cheap package of even falling prices of NVME, seems barely an issue but we really need to see actual UFS pricing. Would you happen to have leads to find prices for these?

UFS does really seem like it makes better sense if Sony can instead use higher powered versions of it for the same heat/consumption budget of an NVME but maybe Sony's too late for this with the upcoming ps5.

Realistically UFS is going to use 0.00001% of a modern CPU, remember its something that runs flawlessly on low-end arm processors. UFS pricing will always be cheaper than NVMe pricing if the 3D Flash NAND is equivalent. Its because they use the same parts with the exception of the controller which always is more expensive on NVMe even on the lowest end models. Like I said the one above which was rated cheaper was only getting 2GBps, if you choose NVMe over UFS, the last thing you would want to do is go cheap and get an underperforming controller because it destroys any reason you had to get NVMe in the first place.

Interestingly enough these patents from Sony from 2015/2016 turned up on a couple of forums. http://www.freepatentsonline.com/y2017/0097897.html

The highlights are that it would use a small cache of SRAM instead of DRAM, it includes a hardware accelerator, and it focuses primarily on high read speeds rather than write speeds very similar to UFS. This goes back to what I said last time.

UFS 3.0 is merely a standard. And the standard was developed to enhance performance in low power scenarios, hence the very low wattage. Sony could easily work with a company to use the same chips but a modified UFS 3.0 controller which doubles power consumption placing in the realm of a low power NVMe and easily exceed 4/5GBps. Alternatively, the controllers are cheap because UFS mostly scales linearly. NVMe has more advanced logic built into the controller hence why they perform better. UFS by contrast is pretty dumb and requires slightly more cpu resources to management. Sony is rumored to have developed an advanced memory logic system to automate this process (Also why NVMe would be redundant). Sony, in theory could drop multiple UFS 3.0 blocks on the board instead. A 128GB block would likely be around $10-$15. They could drop 4 of these, have 512GB of storage with each 128GB block having its own 2.9GBps of bandwidth and let its memory management system sort out the caching.

My proposed off the cuff solution would have yielded a theoretical bandwidth of around 10GBps which is completely unheard of and not even in the near immediate future for PCs, but apparently Sony also worked out their math where much of their patent estimates are based around a 10GBps read speed as well so I guess I wasn't far off.

Also interestingly enough this also addresses the memory logic rumors with its secondary CPU and hardware accelerator built into the controller so its more likely Sony came to the same conclusions and decided to build an SSD controller specific to gaming applications rather than wasting resources of using something that was designed for a different application altogether.

Thanks. It's only sinking in that controllers are what make the difference.

I guess secondary CPUs aren't unheard of? This exists in the PS4 correct?

10GB/s seem insanely fast so if that's gonna be implemented, it's likely to be a small pool cache system correct? I wonder if it's gonna be like 128GB 10 gbps storage + 1gb slower ssd.

Wait. Is UFS an SSD?

Avatar image for michaelmikado
michaelmikado

406

Forum Posts

0

Wiki Points

0

Followers

Reviews: 0

User Lists: 5

#239 michaelmikado
Member since 2019 • 406 Posts

@rzxv04 said:
@michaelmikado said:
@rzxv04 said:

Keep posting. These are all interesting to me.

So UFS might need a slightly better CPU thus lessening some power for games but I guess these are hard to estimate.

I was thinking of high quality video editing built into the PS4 where I only heard that nvme's are taken advantage of copying huge files in and out for editing but I'm not sure how exactly that works. I just remember a youtuber mentioning that the PS4 Pro has horrible quality in their built in capture even compared to the Xbox One S. Was wondering if nvme could help with that but if that were actually an issue at all, UFS might do well enough.

From the looks of it, at least currently, since the controller that's supposed to be expensive is already part of the cheap package of even falling prices of NVME, seems barely an issue but we really need to see actual UFS pricing. Would you happen to have leads to find prices for these?

UFS does really seem like it makes better sense if Sony can instead use higher powered versions of it for the same heat/consumption budget of an NVME but maybe Sony's too late for this with the upcoming ps5.

Realistically UFS is going to use 0.00001% of a modern CPU, remember its something that runs flawlessly on low-end arm processors. UFS pricing will always be cheaper than NVMe pricing if the 3D Flash NAND is equivalent. Its because they use the same parts with the exception of the controller which always is more expensive on NVMe even on the lowest end models. Like I said the one above which was rated cheaper was only getting 2GBps, if you choose NVMe over UFS, the last thing you would want to do is go cheap and get an underperforming controller because it destroys any reason you had to get NVMe in the first place.

Interestingly enough these patents from Sony from 2015/2016 turned up on a couple of forums. http://www.freepatentsonline.com/y2017/0097897.html

The highlights are that it would use a small cache of SRAM instead of DRAM, it includes a hardware accelerator, and it focuses primarily on high read speeds rather than write speeds very similar to UFS. This goes back to what I said last time.

UFS 3.0 is merely a standard. And the standard was developed to enhance performance in low power scenarios, hence the very low wattage. Sony could easily work with a company to use the same chips but a modified UFS 3.0 controller which doubles power consumption placing in the realm of a low power NVMe and easily exceed 4/5GBps. Alternatively, the controllers are cheap because UFS mostly scales linearly. NVMe has more advanced logic built into the controller hence why they perform better. UFS by contrast is pretty dumb and requires slightly more cpu resources to management. Sony is rumored to have developed an advanced memory logic system to automate this process (Also why NVMe would be redundant). Sony, in theory could drop multiple UFS 3.0 blocks on the board instead. A 128GB block would likely be around $10-$15. They could drop 4 of these, have 512GB of storage with each 128GB block having its own 2.9GBps of bandwidth and let its memory management system sort out the caching.

My proposed off the cuff solution would have yielded a theoretical bandwidth of around 10GBps which is completely unheard of and not even in the near immediate future for PCs, but apparently Sony also worked out their math where much of their patent estimates are based around a 10GBps read speed as well so I guess I wasn't far off.

Also interestingly enough this also addresses the memory logic rumors with its secondary CPU and hardware accelerator built into the controller so its more likely Sony came to the same conclusions and decided to build an SSD controller specific to gaming applications rather than wasting resources of using something that was designed for a different application altogether.

Thanks. It's only sinking in that controllers are what make the difference.

I guess secondary CPUs aren't unheard of? This exists in the PS4 correct?

10GB/s seem insanely fast so if that's gonna be implemented, it's likely to be a small pool cache system correct? I wonder if it's gonna be like 128GB 10 gbps storage + 1gb slower ssd.

Wait. Is UFS an SSD?

Yes, a lot of the systems have tiny secondary CPU for various tasks. PS4 had an ARM processor. The thing about a drive that fast is several fold. Let's say PS5 ships with 20GB total RAM. Theoretically you could dump the entirety of your RAM and fill it up in 2 seconds. Traditionally consoles and PCs had large amounts of RAM because pulling data from a disk is too slow. But when you have read speeds this fast you can technically use less high end RAM which is was I mention way way back about extremely fast SSD drives could hypothetically become another tier of RAM. The way the patent is written it seems to claim the SRAM benefits from higher capacities, I would also guess that you would use a smaller drive cache but I'm not sure what they are going for.

Yes, UFS is an SSD. UFS and NVMe are just the interfaces to the exact same Flash NAND or 3D Flash NAND chips. The only difference is how and what they prioritize. That's why it is best to get the solution that benefits the applications you are actually trying to do. In the PS5's case they aren't looking to write a lot of data, but to read a lot of data to have a very fast cache for gaming. Faster than any NVMe drive will ever be in likely the next 3 years. At the expense of write speeds which they don't need but NVMe excels at.

Avatar image for firedrakes
firedrakes

4468

Forum Posts

0

Wiki Points

0

Followers

Reviews: 4

User Lists: 0

#240 firedrakes
Member since 2004 • 4468 Posts

so no one really know what sony or big m is using. where just bs here then..

Avatar image for rzxv04
rzxv04

2578

Forum Posts

0

Wiki Points

0

Followers

Reviews: 0

User Lists: 10

#241 rzxv04
Member since 2018 • 2578 Posts

@michaelmikado said:
@rzxv04 said:
@michaelmikado said:
@rzxv04 said:

Keep posting. These are all interesting to me.

So UFS might need a slightly better CPU thus lessening some power for games but I guess these are hard to estimate.

I was thinking of high quality video editing built into the PS4 where I only heard that nvme's are taken advantage of copying huge files in and out for editing but I'm not sure how exactly that works. I just remember a youtuber mentioning that the PS4 Pro has horrible quality in their built in capture even compared to the Xbox One S. Was wondering if nvme could help with that but if that were actually an issue at all, UFS might do well enough.

From the looks of it, at least currently, since the controller that's supposed to be expensive is already part of the cheap package of even falling prices of NVME, seems barely an issue but we really need to see actual UFS pricing. Would you happen to have leads to find prices for these?

UFS does really seem like it makes better sense if Sony can instead use higher powered versions of it for the same heat/consumption budget of an NVME but maybe Sony's too late for this with the upcoming ps5.

Realistically UFS is going to use 0.00001% of a modern CPU, remember its something that runs flawlessly on low-end arm processors. UFS pricing will always be cheaper than NVMe pricing if the 3D Flash NAND is equivalent. Its because they use the same parts with the exception of the controller which always is more expensive on NVMe even on the lowest end models. Like I said the one above which was rated cheaper was only getting 2GBps, if you choose NVMe over UFS, the last thing you would want to do is go cheap and get an underperforming controller because it destroys any reason you had to get NVMe in the first place.

Interestingly enough these patents from Sony from 2015/2016 turned up on a couple of forums. http://www.freepatentsonline.com/y2017/0097897.html

The highlights are that it would use a small cache of SRAM instead of DRAM, it includes a hardware accelerator, and it focuses primarily on high read speeds rather than write speeds very similar to UFS. This goes back to what I said last time.

UFS 3.0 is merely a standard. And the standard was developed to enhance performance in low power scenarios, hence the very low wattage. Sony could easily work with a company to use the same chips but a modified UFS 3.0 controller which doubles power consumption placing in the realm of a low power NVMe and easily exceed 4/5GBps. Alternatively, the controllers are cheap because UFS mostly scales linearly. NVMe has more advanced logic built into the controller hence why they perform better. UFS by contrast is pretty dumb and requires slightly more cpu resources to management. Sony is rumored to have developed an advanced memory logic system to automate this process (Also why NVMe would be redundant). Sony, in theory could drop multiple UFS 3.0 blocks on the board instead. A 128GB block would likely be around $10-$15. They could drop 4 of these, have 512GB of storage with each 128GB block having its own 2.9GBps of bandwidth and let its memory management system sort out the caching.

My proposed off the cuff solution would have yielded a theoretical bandwidth of around 10GBps which is completely unheard of and not even in the near immediate future for PCs, but apparently Sony also worked out their math where much of their patent estimates are based around a 10GBps read speed as well so I guess I wasn't far off.

Also interestingly enough this also addresses the memory logic rumors with its secondary CPU and hardware accelerator built into the controller so its more likely Sony came to the same conclusions and decided to build an SSD controller specific to gaming applications rather than wasting resources of using something that was designed for a different application altogether.

Thanks. It's only sinking in that controllers are what make the difference.

I guess secondary CPUs aren't unheard of? This exists in the PS4 correct?

10GB/s seem insanely fast so if that's gonna be implemented, it's likely to be a small pool cache system correct? I wonder if it's gonna be like 128GB 10 gbps storage + 1gb slower ssd.

Wait. Is UFS an SSD?

Yes, a lot of the systems have tiny secondary CPU for various tasks. PS4 had an ARM processor. The thing about a drive that fast is several fold. Let's say PS5 ships with 20GB total RAM. Theoretically you could dump the entirety of your RAM and fill it up in 2 seconds. Traditionally consoles and PCs had large amounts of RAM because pulling data from a disk is too slow. But when you have read speeds this fast you can technically use less high end RAM which is was I mention way way back about extremely fast SSD drives could hypothetically become another tier of RAM. The way the patent is written it seems to claim the SRAM benefits from higher capacities, I would also guess that you would use a smaller drive cache but I'm not sure what they are going for.

Yes, UFS is an SSD. UFS and NVMe are just the interfaces to the exact same Flash NAND or 3D Flash NAND chips. The only difference is how and what they prioritize. That's why it is best to get the solution that benefits the applications you are actually trying to do. In the PS5's case they aren't looking to write a lot of data, but to read a lot of data to have a very fast cache for gaming. Faster than any NVMe drive will ever be in likely the next 3 years. At the expense of write speeds which they don't need but NVMe excels at.

Thanks.

Avatar image for zaryia
Zaryia

21607

Forum Posts

0

Wiki Points

0

Followers

Reviews: 0

User Lists: 0

#242 Zaryia
Member since 2016 • 21607 Posts

Well this is clearly a load of bullshit. PS5 will have longer load times than PC, as PS4 did.

Avatar image for mrbojangles25
mrbojangles25

60869

Forum Posts

0

Wiki Points

0

Followers

Reviews: 11

User Lists: 0

#243 mrbojangles25
Member since 2005 • 60869 Posts

@BassMan said:

Even playing games with an NVMe SSD and a powerful CPU on PC does not get rid of load times. So, he is full of shit. Assets and data need to be loaded and processed. That shit takes time and will not be eliminated completely.

Pretty much this.

What I am assuming is that SONY will force every developer to incorporate some sort of minigame or distraction that will make it seem like there is no loading screen. So for example, with God of War you had to push the bridge around and while you were inside doing the thing, the relatively small zones were loading. With other games, they might make some sort of travel segment (say, flying through a hyperspace tunnel or something) where you need to steer your vehicle.

Trying not to be a cynical bastard, but no loading screens? Well, technically no loading screens maybe but there will still be loading involved. Hopefully people don't jump on this and hype it to hell because all it is doing is duping you.

@NoodleFighter said:

Without loading screens how will I get reminder tips from the game!!!???

I kind of like those things, tbh, especially when you can scroll through them. Little tidbits of lore, game tips, and character backgrounds.

Avatar image for michaelmikado
michaelmikado

406

Forum Posts

0

Wiki Points

0

Followers

Reviews: 0

User Lists: 5

#244 michaelmikado
Member since 2019 • 406 Posts

@mrbojangles25:

Load times only exist because disk transfers are slow and assets that may or may not be necessary are loaded into RAM for quick access. It’s literally game development 101.

Avatar image for vfighter
VFighter

11031

Forum Posts

0

Wiki Points

0

Followers

Reviews: 0

User Lists: 0

#245 VFighter
Member since 2016 • 11031 Posts

@zaryia: So every single pc out there has shorter loading times then the ps4? Lol, delusional as always.

Avatar image for ronvalencia
ronvalencia

29612

Forum Posts

0

Wiki Points

0

Followers

Reviews: 1

User Lists: 0

#246  Edited By ronvalencia
Member since 2008 • 29612 Posts

@michaelmikado said:

@rzxv04:

No it would be the opposite, it’s a storage media the idea that the processor would be unable to accept a different storage media that’s literally a semi custom design any is simply untrue.

Further any high end SSD is going to be low volume on the PC side. By contrast UFS is ubiquitous to cellphones, Chromebooks etc. as well so the price dramatically drops over time, again ignoring the other benefits of power consumption and space it is a better option at the targeted limitations of a console. There simply isn’t any benefit to use nvme in a console vs UFS 3.0 which is technically faster than nvme on paper.

https://www.ifixit.com/Teardown/Asus+Chromebook+C202+Teardown/57992

This Intel based Chromebook has SanDisk SDIN9DW416 GB eMMC NAND flash

https://en.wikipedia.org/wiki/Universal_Flash_Storage

Implementation

UFS 2.0 in Snapdragon 820 and 821. Kirin 950 and 955. Exynos 7420

UFS 2.1 in Snapdragon 835, 845 and 850. Kirin 960, 970 and 980. Exynos 9820.

UFS 3.0 in Snapdragon 855. Exynos 9820.[30]

Avatar image for ronvalencia
ronvalencia

29612

Forum Posts

0

Wiki Points

0

Followers

Reviews: 1

User Lists: 0

#247  Edited By ronvalencia
Member since 2008 • 29612 Posts

@michaelmikado said:
@rzxv04 said:

Keep posting. These are all interesting to me.

So UFS might need a slightly better CPU thus lessening some power for games but I guess these are hard to estimate.

I was thinking of high quality video editing built into the PS4 where I only heard that nvme's are taken advantage of copying huge files in and out for editing but I'm not sure how exactly that works. I just remember a youtuber mentioning that the PS4 Pro has horrible quality in their built in capture even compared to the Xbox One S. Was wondering if nvme could help with that but if that were actually an issue at all, UFS might do well enough.

From the looks of it, at least currently, since the controller that's supposed to be expensive is already part of the cheap package of even falling prices of NVME, seems barely an issue but we really need to see actual UFS pricing. Would you happen to have leads to find prices for these?

UFS does really seem like it makes better sense if Sony can instead use higher powered versions of it for the same heat/consumption budget of an NVME but maybe Sony's too late for this with the upcoming ps5.

Realistically UFS is going to use 0.00001% of a modern CPU, remember its something that runs flawlessly on low-end arm processors. UFS pricing will always be cheaper than NVMe pricing if the 3D Flash NAND is equivalent. Its because they use the same parts with the exception of the controller which always is more expensive on NVMe even on the lowest end models. Like I said the one above which was rated cheaper was only getting 2GBps, if you choose NVMe over UFS, the last thing you would want to do is go cheap and get an underperforming controller because it destroys any reason you had to get NVMe in the first place.

Interestingly enough these patents from Sony from 2015/2016 turned up on a couple of forums. http://www.freepatentsonline.com/y2017/0097897.html

The highlights are that it would use a small cache of SRAM instead of DRAM, it includes a hardware accelerator, and it focuses primarily on high read speeds rather than write speeds very similar to UFS. This goes back to what I said last time.

UFS 3.0 is merely a standard. And the standard was developed to enhance performance in low power scenarios, hence the very low wattage. Sony could easily work with a company to use the same chips but a modified UFS 3.0 controller which doubles power consumption placing in the realm of a low power NVMe and easily exceed 4/5GBps. Alternatively, the controllers are cheap because UFS mostly scales linearly. NVMe has more advanced logic built into the controller hence why they perform better. UFS by contrast is pretty dumb and requires slightly more cpu resources to management. Sony is rumored to have developed an advanced memory logic system to automate this process (Also why NVMe would be redundant). Sony, in theory could drop multiple UFS 3.0 blocks on the board instead. A 128GB block would likely be around $10-$15. They could drop 4 of these, have 512GB of storage with each 128GB block having its own 2.9GBps of bandwidth and let its memory management system sort out the caching.

My proposed off the cuff solution would have yielded a theoretical bandwidth of around 10GBps which is completely unheard of and not even in the near immediate future for PCs, but apparently Sony also worked out their math where much of their patent estimates are based around a 10GBps read speed as well so I guess I wasn't far off.

Also interestingly enough this also addresses the memory logic rumors with its secondary CPU and hardware accelerator built into the controller so its more likely Sony came to the same conclusions and decided to build an SSD controller specific to gaming applications rather than wasting resources of using something that was designed for a different application altogether.

Flawed argument since baseline comparison is PS4 with low 60 MB/s to 100 MB/s 1TB HDD 5400 rpm drive.

I copied Reflections_Demo.zip with 1.37 GB size less than 4 seconds with Samsung 840 EVO (750 GB, SATA III) with Intel X299 chipset and AVG AntiVirus real-time scan running in the background.

If 1 GB file takes 4.1 seconds to copy, then there's something wrong with the storage controller.

Avatar image for michaelmikado
michaelmikado

406

Forum Posts

0

Wiki Points

0

Followers

Reviews: 0

User Lists: 5

#248 michaelmikado
Member since 2019 • 406 Posts

@ronvalencia said:
@michaelmikado said:

@rzxv04:

No it would be the opposite, it’s a storage media the idea that the processor would be unable to accept a different storage media that’s literally a semi custom design any is simply untrue.

Further any high end SSD is going to be low volume on the PC side. By contrast UFS is ubiquitous to cellphones, Chromebooks etc. as well so the price dramatically drops over time, again ignoring the other benefits of power consumption and space it is a better option at the targeted limitations of a console. There simply isn’t any benefit to use nvme in a console vs UFS 3.0 which is technically faster than nvme on paper.

https://www.ifixit.com/Teardown/Asus+Chromebook+C202+Teardown/57992

This Intel based Chromebook has SanDisk SDIN9DW416 GB eMMC NAND flash

https://en.wikipedia.org/wiki/Universal_Flash_Storage

Implementation

UFS 2.0 in Snapdragon 820 and 821. Kirin 950 and 955. Exynos 7420

UFS 2.1 in Snapdragon 835, 845 and 850. Kirin 960, 970 and 980. Exynos 9820.

UFS 3.0 in Snapdragon 855. Exynos 9820.[30]

And this one does...

https://www.aboutchromebooks.com/news/cheza-chromebook-with-qualcomm-snapdragon-845-will-have-speedy-ufs-storage/

Avatar image for michaelmikado
michaelmikado

406

Forum Posts

0

Wiki Points

0

Followers

Reviews: 0

User Lists: 5

#249 michaelmikado
Member since 2019 • 406 Posts

@ronvalencia said:
@michaelmikado said:
@rzxv04 said:

Keep posting. These are all interesting to me.

So UFS might need a slightly better CPU thus lessening some power for games but I guess these are hard to estimate.

I was thinking of high quality video editing built into the PS4 where I only heard that nvme's are taken advantage of copying huge files in and out for editing but I'm not sure how exactly that works. I just remember a youtuber mentioning that the PS4 Pro has horrible quality in their built in capture even compared to the Xbox One S. Was wondering if nvme could help with that but if that were actually an issue at all, UFS might do well enough.

From the looks of it, at least currently, since the controller that's supposed to be expensive is already part of the cheap package of even falling prices of NVME, seems barely an issue but we really need to see actual UFS pricing. Would you happen to have leads to find prices for these?

UFS does really seem like it makes better sense if Sony can instead use higher powered versions of it for the same heat/consumption budget of an NVME but maybe Sony's too late for this with the upcoming ps5.

Realistically UFS is going to use 0.00001% of a modern CPU, remember its something that runs flawlessly on low-end arm processors. UFS pricing will always be cheaper than NVMe pricing if the 3D Flash NAND is equivalent. Its because they use the same parts with the exception of the controller which always is more expensive on NVMe even on the lowest end models. Like I said the one above which was rated cheaper was only getting 2GBps, if you choose NVMe over UFS, the last thing you would want to do is go cheap and get an underperforming controller because it destroys any reason you had to get NVMe in the first place.

Interestingly enough these patents from Sony from 2015/2016 turned up on a couple of forums. http://www.freepatentsonline.com/y2017/0097897.html

The highlights are that it would use a small cache of SRAM instead of DRAM, it includes a hardware accelerator, and it focuses primarily on high read speeds rather than write speeds very similar to UFS. This goes back to what I said last time.

UFS 3.0 is merely a standard. And the standard was developed to enhance performance in low power scenarios, hence the very low wattage. Sony could easily work with a company to use the same chips but a modified UFS 3.0 controller which doubles power consumption placing in the realm of a low power NVMe and easily exceed 4/5GBps. Alternatively, the controllers are cheap because UFS mostly scales linearly. NVMe has more advanced logic built into the controller hence why they perform better. UFS by contrast is pretty dumb and requires slightly more cpu resources to management. Sony is rumored to have developed an advanced memory logic system to automate this process (Also why NVMe would be redundant). Sony, in theory could drop multiple UFS 3.0 blocks on the board instead. A 128GB block would likely be around $10-$15. They could drop 4 of these, have 512GB of storage with each 128GB block having its own 2.9GBps of bandwidth and let its memory management system sort out the caching.

My proposed off the cuff solution would have yielded a theoretical bandwidth of around 10GBps which is completely unheard of and not even in the near immediate future for PCs, but apparently Sony also worked out their math where much of their patent estimates are based around a 10GBps read speed as well so I guess I wasn't far off.

Also interestingly enough this also addresses the memory logic rumors with its secondary CPU and hardware accelerator built into the controller so its more likely Sony came to the same conclusions and decided to build an SSD controller specific to gaming applications rather than wasting resources of using something that was designed for a different application altogether.

Flawed argument since baseline comparison is PS4 with low 60 MB/s to 100 MB/s 1TB HDD 5400 rpm drive.

I copied Reflections_Demo.zip with 1.37 GB size less than 4 seconds with Samsung 840 EVO (750 GB, SATA III) with Intel X299 chipset.

If 1 GB file takes 4.1 seconds to copy, then there's something wrong with storage controller.

It’s Sony’s chart not mine, if you have an issue with the methodology of how they measure performance in their own systems which they design and engineer, you’ll have to take it up with them directly.

Avatar image for ronvalencia
ronvalencia

29612

Forum Posts

0

Wiki Points

0

Followers

Reviews: 1

User Lists: 0

#250  Edited By ronvalencia
Member since 2008 • 29612 Posts

@michaelmikado said:
@ronvalencia said:
@michaelmikado said:

@rzxv04:

No it would be the opposite, it’s a storage media the idea that the processor would be unable to accept a different storage media that’s literally a semi custom design any is simply untrue.

Further any high end SSD is going to be low volume on the PC side. By contrast UFS is ubiquitous to cellphones, Chromebooks etc. as well so the price dramatically drops over time, again ignoring the other benefits of power consumption and space it is a better option at the targeted limitations of a console. There simply isn’t any benefit to use nvme in a console vs UFS 3.0 which is technically faster than nvme on paper.

https://www.ifixit.com/Teardown/Asus+Chromebook+C202+Teardown/57992

This Intel based Chromebook has SanDisk SDIN9DW416 GB eMMC NAND flash

https://en.wikipedia.org/wiki/Universal_Flash_Storage

Implementation

UFS 2.0 in Snapdragon 820 and 821. Kirin 950 and 955. Exynos 7420

UFS 2.1 in Snapdragon 835, 845 and 850. Kirin 960, 970 and 980. Exynos 9820.

UFS 3.0 in Snapdragon 855. Exynos 9820.[30]

And this one does...

https://www.aboutchromebooks.com/news/cheza-chromebook-with-qualcomm-snapdragon-845-will-have-speedy-ufs-storage/

That's Qualcomm SoC which is associated with Qualcomm's sphere of influence hardware standards

https://www.jedec.org/news/pressreleases/jedec-publishes-universal-flash-storage-ufs-ufshci-version-30-and-ufs-card

JEDEC's UFS was copied from MIPI Alliance's UFS.

-----------

https://nvmexpress.org/about/nvm-express-overview/

NVMe has "Efficient support for I/O virtualization architectures like SR-IOV"

https://www.amd.com/en/graphics/workstation-virtual-graphics

AMD supports SR-IOV.