Xbox One: Rushed Hardware, Games at 900p, Hardware Failures

This topic is locked from further discussion.

Avatar image for tormentos
tormentos

33793

Forum Posts

0

Wiki Points

0

Followers

Reviews: 0

User Lists: 0

#201  Edited By tormentos
Member since 2003 • 33793 Posts

@ronvalencia said:

@

I don't need a link since TMUs and ROPs are units that consumes memory bandwidth and can be calculated e.g. ROP's and TMU's bytes per cycle rates.

None of the PC hardware has shown a prototype 7850 with 12 CUs (768 stream processors with 48 TMUs) and 153.6GB/s memory bandwidth.

Disabled CUs doesn't change the fundamental processing power of the device i.e. it will remain about 1.3 TFLOPS at 860Mhz from 12 CUs.

For read and write ROPs, 7790's 96 GB/s bandwidth wouldn't even fully use it's 16 ROP's Gpixel rate e.g. 165 GB/s.

http://www.tomshardware.com/reviews/768-shader-pitcairn-review,3196-5.html

If 60 fps was the target frame rate, anything less than 7950 would need to reduce the render resolutions.

Oh yes you need to because you are making arguments based on your opinion,i am making mine based on fact,

The xbox one has a 7790 with 2 CU disable is not a prototype 7850 that card didn't have 14 CU it had 12,the xbox one has 14 with 2 disable,and the regular 7850 has 16 and both 7850 had 32 ROP.

So yeah it is not.

No dude and it has been argue that 2 CU more at 800mhz would have been better than 6% GPU upclock.

The xbox one doesn't have a full 7790 so needing 165Gb/s is even less relevant,even more at resolutions lower than 1080p which is the case with BF4 and several other games,

For 1.18TF you don't need 165GB/s bandwidth or it will serve you for anything,mind you that MS claimed 140GB/s to 150 GB/s that is what you will get no 204GB/s,so yeah even your 165Gb/s number would saturate that banwidth.

In the end i told you so having a weak GPU with high bandwidth will not change anything,BF4 720p just confirmed that.

Avatar image for tormentos
tormentos

33793

Forum Posts

0

Wiki Points

0

Followers

Reviews: 0

User Lists: 0

#202  Edited By tormentos
Member since 2003 • 33793 Posts

@Krelian-co said:

another thread, another 100 posts from ron and the magic secret sauce from xbone

Apparently he can't admit defeat the xbox one can even go sub 720p and he still will not admit it...lol

Avatar image for monsterpuncher
monsterpuncher

177

Forum Posts

0

Wiki Points

0

Followers

Reviews: 0

User Lists: 0

#204  Edited By monsterpuncher
Member since 2013 • 177 Posts

@AtariKidX said:

PS4.......better and cheaper.XBone.......lol

Avatar image for Cherokee_Jack
Cherokee_Jack

32198

Forum Posts

0

Wiki Points

0

Followers

Reviews: 8

User Lists: 2

#205 Cherokee_Jack
Member since 2008 • 32198 Posts
@charizard1605 said:

The console in question [the one that CBOAT is talking about] is a zebra turned stress machine for hardware/heat issues testing.”

Does this make no sense or am I high.

Avatar image for KillzoneSnake
KillzoneSnake

2761

Forum Posts

0

Wiki Points

0

Followers

Reviews: 37

User Lists: 0

#207 KillzoneSnake
Member since 2012 • 2761 Posts

How anyone can buy X1 over PS4 is beyond me. Maybe Ryse low review scores will help them think straight lol

We talking about a more expensive big ugly box that has external power supply and weaker hardware than the smaller more powerful PS4 xD i really dont understand how M$ could make something so horrible. They waste 100 million to make a controller that uses AA batteries xD people better buy that recharge kit separate. Get ready to buy the separate mic adapter too lol.

Avatar image for ronvalencia
ronvalencia

29612

Forum Posts

0

Wiki Points

0

Followers

Reviews: 1

User Lists: 0

#208 ronvalencia
Member since 2008 • 29612 Posts

@@@

@tormentos said:

@ronvalencia said:

@

I don't need a link since TMUs and ROPs are units that consumes memory bandwidth and can be calculated e.g. ROP's and TMU's bytes per cycle rates.

None of the PC hardware has shown a prototype 7850 with 12 CUs (768 stream processors with 48 TMUs) and 153.6GB/s memory bandwidth.

Disabled CUs doesn't change the fundamental processing power of the device i.e. it will remain about 1.3 TFLOPS at 860Mhz from 12 CUs.

For read and write ROPs, 7790's 96 GB/s bandwidth wouldn't even fully use it's 16 ROP's Gpixel rate e.g. 165 GB/s.

http://www.tomshardware.com/reviews/768-shader-pitcairn-review,3196-5.html

If 60 fps was the target frame rate, anything less than 7950 would need to reduce the render resolutions.

Oh yes you need to because you are making arguments based on your opinion,i am making mine based on fact,

The xbox one has a 7790 with 2 CU disable is not a prototype 7850 that card didn't have 14 CU it had 12,the xbox one has 14 with 2 disable,and the regular 7850 has 16 and both 7850 had 32 ROP.

So yeah it is not.

No dude and it has been argue that 2 CU more at 800mhz would have been better than 6% GPU upclock.

The xbox one doesn't have a full 7790 so needing 165Gb/s is even less relevant,even more at resolutions lower than 1080p which is the case with BF4 and several other games,

For 1.18TF you don't need 165GB/s bandwidth or it will serve you for anything,mind you that MS claimed 140GB/s to 150 GB/s that is what you will get no 204GB/s,so yeah even your 165Gb/s number would saturate that banwidth.

In the end i told you so having a weak GPU with high bandwidth will not change anything,BF4 720p just confirmed that.

LOL, You haven't notice X1's has modified it's internal crossbar/bus to support

1. four 64bit channel DDR3.

2. full duplex four 256bit channel ESRAM.

The 7790 has a bottleneck that doesn't enable it use faster memory. Are you so thick that you can't calculate 64 byte per cycle from it's L2 cache?

Your "1.18 TFLOPS you don't need 165 GB/s bandwidth" statement is BS. 165 GB/s comes from ROP's raw read and write requirements NOT CUs. Your mistake.

The prototype 7850 with 12 CUs at 860 Mhz indicates otherwise. Both prototype 7850 with 12 CUs and X1's 12 CUs has similar TMU load/store patterns.

You haven't calculate the raw memory bandwidth to feed the 1.18 TFLOPS.

Each CU's L1 cache has 64byte per cycle rate and there are 12 CUs i.e. that's ~610 GB/s of raw bandwidth. Your second mistake.

The big problem with current GCN design is when workloads overspills beyond CU's SRAM. Outside of the CU, you hit design issues like L2 cache bandwidth and it's memory controller's I/O width and overheads.

GCNs like 7970 has more CUs hence more internal SRAM storage before performance degradation. Lesser GCNs like 7850 encounters performance degradation quicker than 7970.

--------------

Xbox One does NOT have a retail "7790" since it (the 7790) doesn't have X1's memory setup.

7790 = 96 GB/s memory bandwidth from it's 128bit GDDR5. This memory bandwidth is shared between it's TMUs and ROPs.

X1 = 68 GB/s DDR3 and 204 GB/s ESRAM. This memory bandwidth is shared between it's TMUs and ROPs.

Both X1 and 7790's numbers are theoretical values.

--

Lets assume DDR3 ~= GDDR5 in terms of overheads/efficiencies.

X1 = 55 GB/s DDR3 and 150 GB/s ESRAM. 55/68 = 80 precent efficiency.

7790 = 80 precent efficiency on 96 GB/s = 77.64 GB/s.

7770 = 80 precent efficiency on 72 GB/s = 57.6 GB/s.

7850 = 80 precent efficiency on 153.6 GB/s = 122.88 GB/s.

If we use your logic, we'll be claiming R600 codename for Radeon HD 4670 (RV730) LOL.

Disabled CUs does NOT change computation performance i.e. they are inert units. Your PC relative GPU selection doesn't come close to X1's CU (ALUs+TMUs) with memory setup.

What's more important are the following

1. working CUs count since this influences CU's TMU load/store pattern, wavefront buffer size, L1 cache size, LDS size, branch/scalar processor count, total L1 cache byte per cycle rate and 'etc'.

2. closest memory setup to X1.

-------

As for sub-1080p, both consoles has lesser GCN solutions than the 7870 GE replacement i.e. R9-280X.

Both of my 7950 (900Mhz firmware) and 7970 (925 Mhz firmware) are overclocked to 1 Ghz. AMD basically made my 1Ghz overclocked 7970 official i.e. 7970 GE and R9-280X.

Avatar image for ronvalencia
ronvalencia

29612

Forum Posts

0

Wiki Points

0

Followers

Reviews: 1

User Lists: 0

#209  Edited By ronvalencia
Member since 2008 • 29612 Posts

@tormentos said:

@Krelian-co said:

another thread, another 100 posts from ron and the magic secret sauce from xbone

Apparently he can't admit defeat the xbox one can even go sub 720p and he still will not admit it...lol

LOL. You can't admit you can't do simple math.

Unlike you, I look at the small details like bytes per cycle rate in the L1 cache/LDS/L2 cache, wavefront buffer size and 'etc'.

Avatar image for tormentos
tormentos

33793

Forum Posts

0

Wiki Points

0

Followers

Reviews: 0

User Lists: 0

#210  Edited By tormentos
Member since 2003 • 33793 Posts

@ronvalencia said:

@@@

LOL, You haven't notice X1's has modified it's internal crossbar/bus to support

1. four 64bit channel DDR3.

2. full duplex four 256bit channel ESRAM.

The 7790 has a bottleneck that doesn't enable it use faster memory. Are you so thick that you can't calculate 64 byte per cycle from it's L2 cache?

Your "1.18 TFLOPS you don't need 165 GB/s bandwidth" statement is BS. 165 GB/s comes from ROP's raw read and write requirements NOT CUs. Your mistake.

The prototype 7850 with 12 CUs at 860 Mhz indicates otherwise. Both prototype 7850 with 12 CUs and X1's 12 CUs has similar TMU load/store patterns.

You haven't calculate the raw memory bandwidth to feed the 1.18 TFLOPS.

Each CU's L1 cache has 64byte per cycle rate and there are 12 CUs i.e. that's ~610 GB/s of raw bandwidth. Your second mistake.

The big problem with current GCN design is when workloads overspills beyond CU's SRAM. Outside of the CU, you hit design issues like L2 cache bandwidth and it's memory controller's I/O width and overheads.

GCNs like 7970 has more CUs hence more internal SRAM storage before performance degradation. Lesser GCNs like 7850 encounters performance degradation quicker than 7970.

--------------

Xbox One does NOT have a retail "7790" since it (the 7790) doesn't have X1's memory setup.

7790 = 96 GB/s memory bandwidth from it's 128bit GDDR5. This memory bandwidth is shared between it's TMUs and ROPs.

X1 = 68 GB/s DDR3 and 204 GB/s ESRAM. This memory bandwidth is shared between it's TMUs and ROPs.

Both X1 and 7790's numbers are theoretical values.

--

Lets assume DDR3 ~= GDDR5 in terms of overheads/efficiencies.

X1 = 55 GB/s DDR3 and 150 GB/s ESRAM. 55/68 = 80 precent efficiency.

7790 = 80 precent efficiency on 96 GB/s = 77.64 GB/s.

7770 = 80 precent efficiency on 72 GB/s = 57.6 GB/s.

7850 = 80 precent efficiency on 153.6 GB/s = 122.88 GB/s.

If we use your logic, we'll be claiming R600 codename for Radeon HD 4670 (RV730) LOL.

Disabled CUs does NOT change computation performance i.e. they are inert units. Your PC relative GPU selection doesn't come close to X1's CU (ALUs+TMUs) with memory setup.

What's more important are the following

1. working CUs count since this influences CU's TMU load/store pattern, wavefront buffer size, L1 cache size, LDS size, branch/scalar processor count, total L1 cache byte per cycle rate and 'etc'.

2. closest memory setup to X1.

-------

As for sub-1080p, both consoles has lesser GCN solutions than the 7870 GE replacement i.e. R9-280X.

Both of my 7950 (900Mhz firmware) and 7970 (925 Mhz firmware) are overclocked to 1 Ghz. AMD basically made my 1Ghz overclocked 7970 official i.e. 7970 GE and R9-280X.

Hahha............

Dude give it up...

BF4 is 720p on xbox one higher resolution on PS4,and i am sure that quality wise the PS4 will be like medium to high,while the xbox one will be low to medium at best.

The fact that the games is 720p is a testament of under power the xbox one is.

So yeah i told you so no matter all the crap you say 1.18 TF is lower than 7770 performance,you have have 1,000GB/s bandwidth it will change little that the xbox one has a weak GPU.

Giving a 7770 1000GB/s will not make it be as strong as a 7870.

Avatar image for ronvalencia
ronvalencia

29612

Forum Posts

0

Wiki Points

0

Followers

Reviews: 1

User Lists: 0

#211 ronvalencia
Member since 2008 • 29612 Posts

@tormentos said:

@ronvalencia said:

@@@

LOL, You haven't notice X1's has modified it's internal crossbar/bus to support

1. four 64bit channel DDR3.

2. full duplex four 256bit channel ESRAM.

The 7790 has a bottleneck that doesn't enable it use faster memory. Are you so thick that you can't calculate 64 byte per cycle from it's L2 cache?

Your "1.18 TFLOPS you don't need 165 GB/s bandwidth" statement is BS. 165 GB/s comes from ROP's raw read and write requirements NOT CUs. Your mistake.

The prototype 7850 with 12 CUs at 860 Mhz indicates otherwise. Both prototype 7850 with 12 CUs and X1's 12 CUs has similar TMU load/store patterns.

You haven't calculate the raw memory bandwidth to feed the 1.18 TFLOPS.

Each CU's L1 cache has 64byte per cycle rate and there are 12 CUs i.e. that's ~610 GB/s of raw bandwidth. Your second mistake.

The big problem with current GCN design is when workloads overspills beyond CU's SRAM. Outside of the CU, you hit design issues like L2 cache bandwidth and it's memory controller's I/O width and overheads.

GCNs like 7970 has more CUs hence more internal SRAM storage before performance degradation. Lesser GCNs like 7850 encounters performance degradation quicker than 7970.

--------------

Xbox One does NOT have a retail "7790" since it (the 7790) doesn't have X1's memory setup.

7790 = 96 GB/s memory bandwidth from it's 128bit GDDR5. This memory bandwidth is shared between it's TMUs and ROPs.

X1 = 68 GB/s DDR3 and 204 GB/s ESRAM. This memory bandwidth is shared between it's TMUs and ROPs.

Both X1 and 7790's numbers are theoretical values.

--

Lets assume DDR3 ~= GDDR5 in terms of overheads/efficiencies.

X1 = 55 GB/s DDR3 and 150 GB/s ESRAM. 55/68 = 80 precent efficiency.

7790 = 80 precent efficiency on 96 GB/s = 77.64 GB/s.

7770 = 80 precent efficiency on 72 GB/s = 57.6 GB/s.

7850 = 80 precent efficiency on 153.6 GB/s = 122.88 GB/s.

If we use your logic, we'll be claiming R600 codename for Radeon HD 4670 (RV730) LOL.

Disabled CUs does NOT change computation performance i.e. they are inert units. Your PC relative GPU selection doesn't come close to X1's CU (ALUs+TMUs) with memory setup.

What's more important are the following

1. working CUs count since this influences CU's TMU load/store pattern, wavefront buffer size, L1 cache size, LDS size, branch/scalar processor count, total L1 cache byte per cycle rate and 'etc'.

2. closest memory setup to X1.

-------

As for sub-1080p, both consoles has lesser GCN solutions than the 7870 GE replacement i.e. R9-280X.

Both of my 7950 (900Mhz firmware) and 7970 (925 Mhz firmware) are overclocked to 1 Ghz. AMD basically made my 1Ghz overclocked 7970 official i.e. 7970 GE and R9-280X.

Hahha............

Dude give it up...

BF4 is 720p on xbox one higher resolution on PS4,and i am sure that quality wise the PS4 will be like medium to high,while the xbox one will be low to medium at best.

The fact that the games is 720p is a testament of under power the xbox one is.

So yeah i told you so no matter all the crap you say 1.18 TF is lower than 7770 performance,you have have 1,000GB/s bandwidth it will change little that the xbox one has a weak GPU.

Giving a 7770 1000GB/s will not make it be as strong as a 7870.

LOL, 1000 GB/s exceeds the 12 CU's L1 cache bandwidth. hehe you can't do simple math.

I don't care about X1's performance since I have two desktop gaming PCs (i.e. one with 7970 1Ghz and other with 7950 1Ghz), but I will say this... CU alone doesn't equal the entire GPU performance. Each component in the GPU card contributes to the total render time for a frame.

For 7770, L1 cache's 64 byte per cycle x 1000 Mhz x 10 CU = ~596 GB/s

For X1, L1 cache 64 byte per cycle x 853 Mhz x 12 CU = 610 GB/s

---

AMD haven't redesigned their GCNs for very high bandwidth stack memory.

Avatar image for II_Seraphim_II
II_Seraphim_II

20534

Forum Posts

0

Wiki Points

0

Followers

Reviews: 3

User Lists: 0

#212 II_Seraphim_II
Member since 2007 • 20534 Posts

I love that space of time before a new console launches. That's when the fanboy wars are in full swing and people are throwing crap into the air and hoping some of it sticks :P Anyways, my thoughts on the matter, The Xbone will more than likely be less powerful than the PS4 and end up having inferior multiplats. As for the risk of RRPD 2.0. I'm a bit skeptical. After the initial RROD it seems MS went out of its way to avoid anything like that happening again, so I doubt there will be much reliability issues. But if there are, that would be depressing.

Avatar image for darkangel115
darkangel115

4562

Forum Posts

0

Wiki Points

0

Followers

Reviews: 1

User Lists: 0

#213 darkangel115
Member since 2013 • 4562 Posts

@killzowned24 said:

@darkangel115 said:

@killzowned24 said:

PS4 November launch = 32 countries

xbone=13 countries ...lol,no problem at all!!!

most of the PS4 countries have less then 10 consoles available and a price that nobody in those countries can even afford. its all marketing BS. Not to mention some countries don't even have PSN store available to them

LOL, which ones?

United States, Canada, Mexico, Brazil, Australia, New Zealand, United Kingdom, France, Germany, Spain, Italy, Ireland, Austria, Belgium, Denmark, Finland, Luxembourg, Netherlands, Norway, Poland, Portugal, Russia, Sweden, Switzerland, Argentina, Chile, Colombia, Costa Rica, El Salvador, Guatemala, Panama, and Peru.

brazil, argentine, chile, columbia, costa rica, el salvador, guatemala, panama, peru all have extremly low quantities high prices and no PSN support.