• Hey, guest user. Hope you're enjoying NeoGAF! Have you considered registering for an account? Come join us and add your take to the daily discourse.

SCEA files patent for context switching between architecturally distinct GPUs

DieH@rd

Banned
And now more exciting news, SCEA posted a patent application for "DYNAMIC CONTEXT SWITCHING BETWEEN ARCHITECTURALLY DISTINCT GRAPHICS PROCESSORS":

http://appft.uspto.gov/netacgi/nph-...68".PGNR.&OS=DN/20120320068&RS=DN/20120320068
"Graphics processing in a computer graphics apparatus having architecturally dissimilar first and second graphics processing units (GPU) is disclosed. Graphics input is produced in a format having an architecture-neutral display list. One or more instructions in the architecture neutral display list are translated into GPU instructions in an architecture specific format for an active GPU of the first and second GPU."


It seems that they are aiming for APU+GPU combination in PS4. As for architecturaly disctinct graphic processors, they will most likely use 7xxx for APU and 8xxx for discrete GPU... but off course, who knows. Maybe they are just doing preemptive patenting for their patent war chest.

Quoted for the next page.
 

Nirolak

Mrgrgr
I'm splitting this into a separate topic since it's a patent application.

Basically, as mentioned in the OP, this would fit with the rumor that the PlayStation 4 has a distinct, more powerful GPU and also a low power draw (presumably Laptop mobile?) GPU that is part of its processor setup (an AMD APU).

It's actually stated right here in the patent:

Patent said:
5. The apparatus of claim 2 wherein the first GPU is a high power GPU and the second GPU is a low power GPU having lower power consumption than the high power GPU and a maximum processing capacity that is less than a maximum processing capacity of the high power GPU.

6. The apparatus of claim 5 wherein the just-in-time compiler is configured to perform a context switch from the high power GPU to the low power GPU if the high power GPU is the active GPU and the high power GPU is operating at a processing capacity that is less than or equal to the maximum processing capacity of the low power GPU.

7. The apparatus of claim 5 wherein the just-in-time compiler is configured to perform a context switch from the low power GPU to the high power GPU if the low power GPU is the active GPU, the low power GPU is operating at its maximum processing capacity, and a frame render time for the apparatus is decreasing.

Edit:

The proposed system also has video RAM.

Patent said:
12. The method of claim 11, wherein performing the context switch further comprises transferring contents of a video RAM of the GPU that is inactive after the context switch to a video RAM of the GPU that is to be active after the context switch.

What the second GPU is used for is debatable, but may be detailed in the long patent. I'm skimming through it more as I have time.

Edit 2:

The more I read of this, the more it sounds like it's for power saving (like in Apple laptops) and they might not be using the second GPU as a coprocessor.

Patent said:
[0038] As noted above, only one of the GPU 304A, 304B is active at a time. The active GPU may periodically output pixel data for an image from the corresponding graphics memory to be displayed on the display device 310. The display device 308 may be any device capable of displaying visual information in response to a signal from the client device 300, including CRT, LCD, plasma, and OLED displays. The display controller 308 may convert the pixel data to signals that display device 310 uses to generate visible images. The display controller 308 may provide the display device 310 with analog or digital signals. By way of example, the display 310 may include a cathode ray tube (CRT) or flat panel screen that displays visible text, numerals, graphical symbols or images.

They do note this though:

Patent said:
[0046] Embodiments of the present invention as described herein may be extended to enable dynamic load balancing between two or more graphics processors for the purpose of increasing performance at the cost of power, but with architecturally similar GPUs (not identical GPUs as with SLI). By way of example, and not by way of limitation, a context switch may be performed between the two similar GPUs based on which one would have the higher performance for processing a given set of GPU input. Performance may be based, e.g., on an estimated amount of time or number of processor cycles to process the input.
 
Does architecturally distinct mean that they're using actual different architectures for the two GPUs or just that one is dedicated and one is on the same die as the CPU?
 

mrklaw

MrArseFace
I'm splitting this into a separate topic since it's an actual patent.

Basically, as mentioned in the OP, this would fit with the rumor that the PlayStation 4 has a distinct desktop GPU and also a low power draw mobile GPU that is part of its processor setup (an AMD APU).

It's actually stated right here in the patent:

*if* the patent is something they are bringing to market and not just something they filed in the normal course of research etc,

And would switching also suggest either/or, so low power GPU when running Netflix for example, and big GPU for games? So the smaller GPU might not be usable in games to add processing capacity?
 
Not a bad idea when you think about it. Most moblie-class GPUs can easily handle things like rendering a UI or playing back HD video. Makes sense to disable the thing you'll be using for games when that extra power isn't needed.
 

Nirolak

Mrgrgr
So in layman's terms...is this good or bad for PS4?

Good, since it means you get the power of the main GPU + the mobile GPU.*

It also means they definitely have a higher powered GPU and a lower powered GPU, implying it's not just what is in the APU.

Assuming they're using the technology in the patent of course.

*Dependent on the main GPU being good, and this actually being used directly in game processing.
 

Kuro Madoushi

Unconfirmed Member
I'm splitting this into a separate topic since it's an actual patent.

Basically, as mentioned in the OP, this would fit with the rumor that the PlayStation 4 has a distinct desktop GPU and also a low power draw mobile GPU that is part of its processor setup (an AMD APU).

It's actually stated right here in the patent:
Just to be clear, it is a patent application NOT a registered patent so it hasn't been examined yet.

Still an interesting concept though if it is for PS4
 
Sounds like they could use one GPU/CPU for physics and ai related stuff and one purely for graphics. This sounds like it could be a beast of a console if I'm interpreting this right.
 
Good, since it means you get the power of the main GPU + the mobile GPU.*

It also means they definitely have a primary GPU and aren't just using a mobile GPU.

Assuming they're using the technology in the patent of course.

*Dependent on the main GPU being good.

Hm alright, I guess that's one way to look at it.

Thanks for the info :)
 

ascii42

Member
Good, since it means you get the power of the main GPU + the mobile GPU.

It also means they definitely have a primary GPU and aren't just using a mobile GPU.

Assuming they're using the technology in the patent of course.

Couldn't they be doing mobile+integrated like some laptops, or are both graphics chips discrete in this patent?
 

Nirolak

Mrgrgr
Just to be clear, it is a patent application NOT a registered patent so it hasn't been examined yet.

Still an interesting concept though if it is for PS4

True, though I feel that has more impact on Sony's ability to block others than indicating their intentions aren't in this direction.
 

i-Lo

Member
Pertaining to context switching, can anyone make me understand what it is? As in does it have to do with task graphical task switching between APU and GPU in real time?

So in layman's terms...is this good or bad for PS4?

It's something we all knew as a rumour TBH from the beginning, i.e., APU + discrete GPU set up. It means that the PS4 will consume little power when performing all non-gaming based functions.
 

sinxtanx

Member
*if* the patent is something they are bringing to market and not just something they filed in the normal course of research etc,

And would switching also suggest either/or, so low power GPU when running Netflix for example, and big GPU for games? So the smaller GPU might not be usable in games to add processing capacity?

Actually it's probably there for actually processing games. Compute shaders are getting pretty big these days.
 

xero273

Member
is this the same thing like how it works on macbooks? on the desktop the laptop uses the intel gpu and games it would switch to the nvidia video card
 

i-Lo

Member
It says, "video ram". So if PS4 is to have above 4GB of RAM, it would mean they aren't going to the unified solution.
 

Nirolak

Mrgrgr
*if* the patent is something they are bringing to market and not just something they filed in the normal course of research etc,

And would switching also suggest either/or, so low power GPU when running Netflix for example, and big GPU for games? So the smaller GPU might not be usable in games to add processing capacity?

I edited my post to reflect this possibility since I didn't read the whole thing before splitting the thread admittedly.

I'll edit again when I get through it all, though that might be a bit since I'm at work.
 

ghst

thanks for the laugh
Good, since it means you get the power of the main GPU + the mobile GPU.*

It also means they definitely have a primary GPU and aren't just using a mobile GPU.

does it? mobile gpus are discrete chips. there's nothing in the article to suggest that it's not going have a separate apu for low power "desktop mode" equivalent + a mobile gpu for gaming. plenty of laptops already do this.
 

Nirolak

Mrgrgr
does it? mobile gpus are discrete chips. there's nothing in the article to suggest that it's not going have a separate apu for low power "desktop mode" equivalent + a mobile gpu for gaming. plenty of laptops already do this.

Okay, so, I guess it could be a higher powered mobile GPU, but it does specifically list that one of these is more powerful than the other.

There, edited the explanation post to reflect "more powerful gpu" and "less powerful gpu".
 
One could also be for watching blu-rays, netflix, amazon, and other media services while the other is for games. This would reduce power consumption and not put such a strain on the console. They could activate both in games if they wanted probably for additional power.
 

i-Lo

Member
What would they need a 2nd low power gpu for? For gui visualization?

All non-gaming related function. It would consume very little power compared to its peak ability.

Makes me wonder if the system will also have the same power envelope as PS3 ~210W..
 
What would they need a 2nd low power gpu for? For gui visualization?

Background services. GPU based encoding for a DVR function is one example. It would allow Sony to keep a lot of reserve power, this console needs to last until 2020 and stay relevant, it means they need to account for as yet non-existent technical innovations. Keeping power in reserve by way of an on die GPU is sensible.
 

USC-fan

Banned
Its not going to be a mobile parts since those are binned parts. Its going to be a custom built apu based on Kaveri, Kabini, or Temash designs. Its going to take advantages of amd hsa features. Makes perfect sense...

Kaveri is the high power apu and the gpu itself on that chip is going to be over 1 tflop.
 

AgentP

Thinks mods influence posters politics. Promoted to QAnon Editor.
So in layman's terms...is this good or bad for PS4?

Good, you typically don't spend money to patent bad ideas. It is just like what laptops with discrete GPUs do. For low power desktop stuff they use the CPU's video pipeline and when they need real power it switches to the discrete Nvidia GPU.
 
Hopefully that means that they've added multi-tasking background services. I would love to pause the game, go to a PSN Store or Browser and then come back. That or go to Netflix and return to the game. Multiple services running at the same time would be fantastic! Dedicate the bigger GPU for gaming and smaller GPU for everything else.

In simple words: Similar to how Vita or any smartphone works, but more Steam-like.
 

Elios83

Member
More likely to be bad than good. If it were something non-pathetic in there e.g. a 7870/8870 or better then it'd be highly unlikely that they'd stick an APU in there as well.

That hasn't a lot of sense, if they are concerned that the main GPU is weak they wouldn't add a separate GPU in an other chip to compensate but they would simply choose a more powerful main GPU.
The GPU in the APU adds a lot of flexibility.
It can be used as a support for the CPU (which is in the same APU) using GPGPU, it could be used to help the main GPU at specific tasks, it could be used to add multiple video outputs and/or 3D applications multitasking, it can be used to run specific low power applications (like simple games, emulators of old consoles).
So it's definetly a great thing also because using APU+GPU gives many options for the memory system.
 

Nirolak

Mrgrgr
Reading through this more it doesn't seem like they're combining the output and are instead doing this for power saving.

Patent said:
[0038] As noted above, only one of the GPU 304A, 304B is active at a time. The active GPU may periodically output pixel data for an image from the corresponding graphics memory to be displayed on the display device 310. The display device 308 may be any device capable of displaying visual information in response to a signal from the client device 300, including CRT, LCD, plasma, and OLED displays. The display controller 308 may convert the pixel data to signals that display device 310 uses to generate visible images. The display controller 308 may provide the display device 310 with analog or digital signals. By way of example, the display 310 may include a cathode ray tube (CRT) or flat panel screen that displays visible text, numerals, graphical symbols or images.

That said, they do state this:

Patent said:
[0046] Embodiments of the present invention as described herein may be extended to enable dynamic load balancing between two or more graphics processors for the purpose of increasing performance at the cost of power, but with architecturally similar GPUs (not identical GPUs as with SLI). By way of example, and not by way of limitation, a context switch may be performed between the two similar GPUs based on which one would have the higher performance for processing a given set of GPU input. Performance may be based, e.g., on an estimated amount of time or number of processor cycles to process the input.
 

Clear

CliffyB's Cock Holster
It just looks like a slightly different take on the load-balancing mechanism that MS has been talking about. OS alone uses the low-spec hardware config, while games and power apps use the full spec.

The idea is that with 2 sets of VRAM you can keep your frame-buffers intact whilst working on output with the support GPU, such as doing OS stuff whilst in the middle of gaming session.
You dont really need multiple CPU's (which I believe was MS approach) because you can simply suspend processing of the task that is no longer on-screen.
 

androvsky

Member
Aren't the 7xxx and 8xxx series from AMD supposed to be pretty similar from a straight architecture point of view? I think it's a bit odd Sony would be using two GPUs from different generations no matter how similar, but I guess they could be doing a bit of bargain shopping to cut down on engineering costs.

It'd be hilarious if they were thinking about using the RSX for their APU, but it's a bad pick for a ton of reasons (no compute shaders, no built-in video decoder, probably not as low power as a similar new GPU would be).
 
One could also be for watching blu-rays, netflix, amazon, and other media services while the other is for games. This would reduce power consumption and not put such a strain on the console. They could activate both in games if they wanted probably for additional power.

Doesn't that mean that they will underpay on the CPU though? they seriously are going to trade CPU performance for power consumption? sounds extremely stupid If I'm understanding this correctly.
 

USC-fan

Banned
Aren't the 7xxx and 8xxx series from AMD supposed to be pretty similar from a straight architecture point of view? I think it's a bit odd Sony would be using two GPUs from different generations no matter how similar, but I guess they could be doing a bit of bargain shopping to cut down on engineering costs.

It'd be hilarious if they were thinking about using the RSX for their APU, but it's a bad pick for a ton of reasons (no compute shaders, no built-in video decoder, probably not as low power as a similar new GPU would be).

Amd apu design are always one Gen behind the gpus. This is good news since it means thy are going with sea island gpu.

Amd apu must be pair with the match gpu. This means they are design around that and found a solution that fixes that problem.
 

Elios83

Member
Doesn't that mean that they will underpay on the CPU though? seriously are they trading CPU performance for power consumption? sounds extremely stupid If I'm understanding this correctly.

No the point is to enhance the CPU power through GPGPU which also adds a lot of flexibility to the process because it allows for the same kind of multitasking only possible with multiple GPUs.
Basically it's like having a part of the CPU which is very flexible and can be used in a lot of ways for different applications. In my mind it's a pretty great thing.
 

KageMaru

Member
It just looks like a slightly different take on the load-balancing mechanism that MS has been talking about. OS alone uses the low-spec hardware config, while games and power apps use the full spec.

The idea is that with 2 sets of VRAM you can keep your frame-buffers intact whilst working on output with the support GPU, such as doing OS stuff whilst in the middle of gaming session.
You dont really need multiple CPU's (which I believe was MS approach) because you can simply suspend processing of the task that is no longer on-screen.

This is exactly what I was thinking reading through this.
 

Globox_82

Banned
Can someone answer on this question.

If this is true and Sony ends up using this patent, what does it mean really, that MS can't have APU + GPU (low end gpu + powerful gpu)? Aren't both going for that? What is this patent making that only Sony can use?


and even more importantly, let's say low power GPU is 0.5 tflops and main is 2tflops, can those combined 2.5tflops be used for gaming or only 2tflops?
 
Hard to determine wether this is something for a Vaio laptop, a Sony Tablet, Sony Phone, or even a Gaikai server park for cloud gaming rather than PS4 though ... but you never know. But the abstraction layer sounds like it would be more suitable for something Gaikai or mobile related than PS4 perhaps?
 
Top Bottom