Discrete PC CPUs and GPUs on life support, Nvidia and Intel in grave danger! Get in now!

Putting Ryzen cores on the APU would take up much more room than Jaguar cores do currently, hence you can't just put a full fat Vega GPU and Ryzen together, die size would become enormous and the costs would be too much, discrete parts for high end performance makes much more sense

Plus cooling that APU with high end Ryzen(6 to 8 core) and Vega(with more than 8 CU's) on one die would be very tough, would need at least a water cooler to keep it cool

Ever heard of the X1X? They managed a 6 teraflop GPU and sure the CPU is dated but if it released next year it would certainly be Ryzen based and for the same $500. Ok, now we're talking about the almighty PC, you are telling me they can't even match the console peasants?

It's just a matter of time before they introduce gaming APUs for Desktop PCs. It's amazingly dumb of them that they have been already doing it for consoles for years but until now it hasn't occurred to them to bring it over to the PC.

Imagine gaming PCs the size of an ultrabook but with the power of a Ryzen CPU and a Vega GPU. This really could put PCs in the living room as the ultimate media hub. This could be the true steambox as well at the same time.

Extinct will be the monster PC tower with tower CPU cooler and massive GPU. Instead you will get a smaller, energy efficient truly small form factor PC with actual all around power including good gaming performance.

Laptops will benefit greatly as well because gaming performance will be phenomenal so it makes sense this is where they begin.

Now what about Intel and Nvidia? How will intel compete if they don't have real graphics on their CPU? How will Nvidia compete if they don't have a CPU with their graphics? An AMD APU can kill both Intel and Nvidia in short order under the right circumstances.

The future is a small Apple TV size device with high end PC like performance. I guess building PCs may even become obsolete at some point as well.

No, just,,,just NO!

Ever heard of the X1X? They managed a 6 teraflop GPU and sure the CPU is dated but if it released next year it would certainly be Ryzen based and for the same $500. Ok, now we're talking about the almighty PC, you are telling me they can't even match the console peasants?

Ever heard of the X1X? They managed a 6 teraflop GPU and sure the CPU is dated but if it released next year it would certainly be Ryzen based and for the same $500. Ok, now we're talking about the almighty PC, you are telling me they can't even match the console peasants?

Do you ignore the fact that X1X is using a desktop sized gpu cooler to cool the APU which is as large as full sized Polaris gpu+ cpu die using a 245w PSU to power the X1X.... VS putting a Ryzen APU in a laptop that is as strong ..... yea its not going to happen.... Purpose of an APU is to simplify manufacturing of the motherboard where the design don't need an integrated graphics gpu soldered on the motherboard or using a discrete graphics card. Also an advantage in using a cpu+gpu combo under one die "APU" is that its improves power efficiency by allowing it to share resources.

There is a difference between the design of a customized box ie console which allows much more headroom vs the limits and market of mobile and low powered segment of computers using APU's.

Intel/AMD's new mobile SOC 45 watts from https://wccftech.com/intel-kaby-lake-g-series-integrated-radeon-gpus-first-benchmarks-specifications/ has a GPU with 24 CU at 1190 Mhz (3.66 TFLOPS) and single stack HBM v2 VRAM at 800Mhz (~205 GB/s memory bandwidth).

Unlike PS4 Pro's APU, Intel/AMD's new SOC wouldn't be gimped by crappy standard Jaguar CPUs which is very important for 1080p/1440p resolution.

From your link:

"gfx804", so a Polaris GPU?

"Discrete GPU" means a GPU with discrete video memory.

Ron Valencia

Wrong. That's a singular aspect of the definition.

A discrete GPU is a Graphics Processing Unit that is not part of the CPU, ie, it's a separate add in board or a separate chip on the system board (usually in laptops and AIOs.) Discrete GPUs also use their own VRAM instead of using system RAM for video processing.

Intel/AMD's new SoC(System on a Chip)'s GPU has it's own VRAM.

Too bad for you, Intel/AMD's new SoC's 24 CU GPU has beaten NVIDIA MX150/1030/1050/1050 Ti and AMD RX-560 with 16 CU.

https://en.wikipedia.org/wiki/Graphics_processing_unit#Dedicated_graphics_cards

A dedicated GPU is not necessarily removable, nor does it necessarily interface with the motherboard in a standard fashion. The term "dedicated" refers to the fact that dedicated graphics cards have RAM that is dedicated to the card's use, not to the fact that most dedicated GPUs are removable. Further, this RAM is usually specially selected for the expected serial workload of the graphics card (see GDDR).

And? That does not make it a discrete GPU because it fails the first half of the definition; you know the more important portion of the defintion that precedes the "also" portion.

The term "dedicated GPU" is entirely different than "discrete GPU."

NVIDIA's product line does not end at the 1050 line.

Glad we could clear that up.

Your definition was created by a non-computer engineer.

The purpose for Intel/AMD SOC was to reduce power consumption and reduce board size while beating NVIDIA MX150/1030/1050/1050 Ti and AMD RX-560 with 16 CU.

Laptop PCs has more than 50 percent of PC market and it's the growing part of the business.

Putting Ryzen cores on the APU would take up much more room than Jaguar cores do currently, hence you can't just put a full fat Vega GPU and Ryzen together, die size would become enormous and the costs would be too much, discrete parts for high end performance makes much more sense

Plus cooling that APU with high end Ryzen(6 to 8 core) and Vega(with more than 8 CU's) on one die would be very tough, would need at least a water cooler to keep it cool

Do you ignore the fact that X1X is using a desktop sized gpu cooler to cool the APU which is as large as full sized Polaris gpu+ cpu die using a 245w PSU to power the X1X.... VS putting a Ryzen APU in a laptop that is as strong ..... yea its not going to happen.... Purpose of an APU is to simplify manufacturing of the motherboard where the design don't need an integrated graphics gpu soldered on the motherboard or using a discrete graphics card. Also an advantage in using a cpu+gpu combo under one die "APU" is that its improves power efficiency by allowing it to share resources.

There is a difference between the design of a customized box ie console which allows much more headroom vs the limits and market of mobile and low powered segment of computers using APU's.

Wow, I wasn't talking about laptops being capable of replacing high end gaming PCs. What I was referring to was the natural progression of things. AMD builds the laptop APU for the laptop space and then eventually they decide to build a high end APU for desktops that would be similar to the X1X APU but a lot better.

If AMD released an APU with the power of an RX 580 GPU and the power of a Ryzen 4 core/8 thread CPU for say $399, I think it would sell like hotcakes. I could be wrong but that's what I think. No it wouldn't be more powerful than a high end PC with an i7 8700k and GTX 1080 ti, but it would potentially fit in a PC the size of a slim blu-ray player or maybe Xbox One X console and be inexpensive enough and good enough that it would sell like crazy.

Sure thing. You might want to correct the folks at PC magazine or everywhere else on the web who seem to agree on a definition that differs significantly from yours.

I don't know what point you believe that you're making by continuing to bring up the Intel/AMD SoC since it's still beaten by NVIDIA's higher end products.

I have no idea why you're mentioning laptop sales either. Laptop PCs are pre-built (often closed) systems, so claiming that they're 50% of the PC market means nothing when the majority of people that game on PCs will still prefer to build their own from separate components because they can produce more powerful systems.

The people purchasing laptops are more likely to be purchasing them for portability and productivity not power.

Useless definition for MCM (multi-chip-modules). The core idea from "discrete GPU" is a GPU having it's own discrete VRAM.

"NVIDIA's higher end products" has higher TDP and the mobility factor is less.

Anyway, https://www.notebookcheck.net/NVIDIA-GeForce-GTX-1050-Notebook.178614.0.html

GTX 1050 mobile has 40 to 50 watts.

Again, the purpose for Intel/AMD SOC is to reduce power consumption and reduce board size while beating NVIDIA MX150/1030/1050/1050 Ti and AMD RX-560 with 16 CU.

APUs up until now have failed miserably because they have been very flawed. AMD APUs have had a weak CPU and weak GPU so that makes them practically useless. No way anyone is replacing a high end PC with that. Intel CPUs have graphics but terrible performance. You can't do anything with only an Intel CPU, you have to pair it with a powerful GPU to have good gaming performance.

This is the first time we have seen a Ryzen level CPU and Vega level GPU on one chip. This could replace a high end PC and be a very small form factor. This is the future of PCs and could be AMD's chance to really take command of the industry as they have already done in the console space.

And? That does not make it a discrete GPU because it fails the first half of the definition; you know the more important portion of the defintion that precedes the "also" portion.

The term "dedicated GPU" is entirely different than "discrete GPU."

NVIDIA's product line does not end at the 1050 line.

Glad we could clear that up.

Actually no, it passes the first half, though in a somewhat roundabout way. It's a separate chip from the CPU.

A system on a chip or system on chip (SoC or SOC) is an integrated circuit (also known as an "IC" or "chip") that integrates all components of a computer or other electronic systems. It may contain digital, analog, mixed-signal, and often radio-frequency functions—all on a single substrate

The last part is important and is why it does not fit the traditional definition. It's all part of the same silicon, but its functionality is separated while simultaneously being on the same "chip."

It's the reason they have to rely on lower power draw to prevent the whole thing from melting. This means that its processing capabilities must also be restrained in order to even achieve these integrated graphic advances.

Compromise has to be made and that's why actual discrete components remain > than integrated solutions unless they can figure out an end-around the laws of physics as we currently understand them.

Discrete - individually separate and distinct.

And it's not a "useless" definition. It is the established definition for discrete GPU implying both a physical separation from the processor as well as separate memory. Under the current definition of the term, it would be a misnomer to refer to the GPU portion of an SoC as a "discrete GPU." It would be more accurately described as an on-die GPU with discrete memory. This <> a "discrete GPU" under the existing definition.

As far as NVIDIA's higher end products having higher TDP and less mobility, duh?

Reducing power consumption is a necessity with these chips because they would otherwise fry. I'm not diminishing Intel/AMD's achievement, but it's not as though NVIDIA has ceased development and will be eclipsed by these products any time soon.

Integrated solutions have improved dramatically in the last 5+ years, but I see no reason to believe traditional GPUs will not continue to do the same.

@ronvalencia said:

@Dasein808 said:

@ronvalencia said:

Useless definition for MCM (multi-chip-modules). The core idea from "discrete GPU" is a GPU having it's own discrete VRAM.

"NVIDIA's higher end products" has higher TDP and the mobility factor is less.

Anyway, https://www.notebookcheck.net/NVIDIA-GeForce-GTX-1050-Notebook.178614.0.html

GTX 1050 mobile has 40 to 50 watts.

Again, the purpose for Intel/AMD SOC is to reduce power consumption and reduce board size while beating NVIDIA MX150/1030/1050/1050 Ti and AMD RX-560 with 16 CU.

Discrete - individually separate and distinct.

And it's not a "useless" definition. It is the established definition for discrete GPU implying both a physical separation from the processor as well as separate memory. Under the current definition of the term, it would be a misnomer to refer to the GPU portion of an SoC as a "discrete GPU." It would be more accurately described as an on-die GPU with discrete memory. This <> a "discrete GPU" under the existing definition.

As far as NVIDIA's higher end products having higher TDP and less mobility, duh?

Reducing power consumption is a necessity with these chips because they would otherwise fry. I'm not diminishing Intel/AMD's achievement, but it's not as though NVIDIA has ceased development and will be eclipsed by these products any time soon.

Integrated solutions have improved dramatically in the last 5+ years, but I see no reason to believe traditional GPUs will not continue to do the same.

So if this chart is accurate, since it's leaked benchmarks for now, all this means is that the ultrabook market will have some options for "decent" gaming performance finally.

This will not affect the desktop market whatsoever, and I'm sure solutions like this will carry a price premium based on performance/size. I think TC actually believes that these APUs won't carry any sort of price premium, and you're getting the included GPU for free.

@ronvalencia said:

So if this chart is accurate, since it's leaked benchmarks for now, all this means is that the ultrabook market will have some options for "decent" gaming performance finally.

This will not affect the desktop market whatsoever, and I'm sure solutions like this will carry a price premium based on performance/size. I think TC actually believes that these APUs won't carry any sort of price premium, and you're getting the included GPU for free.

A system on a chip or system on chip (SoC or SOC) is an integrated circuit (also known as an "IC" or "chip") that integrates all components of a computer or other electronic systems. It may contain digital, analog, mixed-signal, and often radio-frequency functions—all on a single substrate

The last part is important and is why it does not fit the traditional definition. It's all part of the same silicon, but its functionality is separated while simultaneously being on the same "chip."

It's the reason they have to rely on lower power draw to prevent the whole thing from melting. This means that its processing capabilities must also be restrained in order to even achieve these integrated graphic advances.

Compromise has to be made and that's why actual discrete components remain > than integrated solutions unless they can figure out an end-around the laws of physics as we currently understand them.