Current Generation Hardware Speculation with a Technical Spin [post GDC 2020] [XBSX, PS5]

[Frenetic Pony]

Mesh shading should help improve saturation across a wider chip also.

Still can't believe Sony didn't go for mesh shaders. And a lot of the press briefing just felt like Cerny trying to make up for realizing MS beat him in hardware design this time around. No mesh shaders, no variable rate shading (hell a 10% perf boost with little visible change ala Wolfenstein is still 10%), a bad bandwidth to compute ratio, etc. He did get his little custom sound chip on there. That thing probably cost what, a hundred million in design?

A 44cu GPU (just salvage from 48) with high clockspeeds from actual RDNA2 and moving to 16gbps GDDR6 could've put it nigh even with MS in a smaller package. And if you want something custom why wouldn't you go for raytracing? That's the big shiny thing that's obvious when people watch videos, double the raytracing units from a normal gpu and have an ultra high bandwidth 128mb ESRAM cache so you don't go out to main memory when going through the bvh. Or, something anyway. Ohwell.

 

[Betanumerical]

Still can't believe Sony didn't go for mesh shaders. And a lot of the press briefing just felt like Cerny trying to make up for realizing MS beat him in hardware design this time around. No mesh shaders, no variable rate shading (hell a 10% perf boost with little visible change ala Wolfenstein is still 10%), a bad bandwidth to compute ratio, etc. He did get his little custom sound chip on there. That thing probably cost what, a hundred million in design?

A 44cu GPU (just salvage from 48) with high clockspeeds from actual RDNA2 and moving to 16gbps GDDR6 could've put it nigh even with MS in a smaller package. And if you want something custom why wouldn't you go for raytracing? That's the big shiny thing that's obvious when people watch videos, double the raytracing units from a normal gpu and have an ultra high bandwidth 128mb ESRAM cache so you don't go out to main memory when going through the bvh. Or, something anyway. Ohwell.

I wasn't aware that we knew the PS5 didn't have VRS or mesh shaders?.

 

[Frenetic Pony]

I wasn't aware that we knew the PS5 didn't have VRS or mesh shaders?.

There was no mention of VRS, and there was a mention of primitive shaders, a precursor from RDNA1 that's non programmable, at least in consumer RDNA chips.

 

[blakjedi]

Unlikely, at 10% CPU and 10% GPU utilization, I would expect both to run at 3.5 and 2.223 respectively. At some utilization, they would each start to downclock. If the CPU is at a lower utilization, and the GPU is reaching its threshold, then smartshift can kick in and provide additional power to the GPU to avoid downclocking as soon. I question Alex's notion of developers actually having to choose a speed setting. It seems much more straight forward to simply allow the system to adjust dynamically in realtime and optimize performance in given scenes as normal. The only drawback there is mostly on the CPU side where certain game systems may expect a fixed clock. Perhaps that's the element they allow the developer to define.

So then these are neither peak nor boost clock speeds as we know them.

If the clocks are absolutely 2.23 and 3.5... then whats the use of smartshift?

What power could be "sent" to the GPU if its already at full 2.23GHZ?

This doesn't seem like a real thing..

 

[Scott_Arm]

There was no mention of VRS, and there was a mention of primitive shaders, a precursor from RDNA1 that's non programmable, at least in consumer RDNA chips.

primitive shaders in RDNA2 are likely the hardware that underpins the Mesh Shader API that microsoft is providing. Also, not mentioning something doesn't mean it isn't there. It's still very possible the GPU features VRS hardware because it's an RDNA2-based GPU.

 

[3dilettante]

Mesh shading should help improve saturation across a wider chip also.

That should be true, and that may come with additional CU wavefronts for whatever shader is used for the geometry processing, since it's intended to alleviate a bottleneck that would have constrained the amount of geometry work that would be done at any given time.

primitive shaders in RDNA2 are likely the hardware that underpins the Mesh Shader API that microsoft is providing. Also, not mentioning something doesn't mean it isn't there. It's still very possible the GPU features VRS hardware because it's an RDNA2-based GPU.

Cerny touched on some use cases that would appear to be some kind of Primitive Shader version 2, although it was too brief a mention to pin down if retaining the name Primitive Shader means there are differences in features or custom elements.
Just by offering programmer access to them, Sony is implying that some kind of enhancement work was done since whatever version was in Vega was not an acceptable offering. Driver changes hint at RDNA replacing or modifying system hooks that Vega had for the NGG pipeline and primitive shaders, so hopefully this new version is more approachable than the culling-only original.

 

[jlippo]

But PS4pro had optimizations like Rapid Packed Math and Checkerboard Rendering that were absent on Xbox One X. We know that Series X supports math all the way down to 4bit, and can assume that PS5 supports at least fp16 because of hardware backwards compatibility, and probably checkerboard rendering as well. But based on what happened this generation, where checkerboarding seams to have been either ignored in favor of other reconstruction techniques. Or perhaps used on PS4 and another solution used on One X that is comparable, making CBR less of an advantage. There doesn't seam to be that secret sauce in this upcoming generation that can help Sony level the playing field. In fact, Sony has made no mention of things like VRS that could really swing performance in MS favor if Sony doesn't have it, and it's easy to implement.

I wonder if MS's support of multiple FP formats (FP4/FP8/FP16/FP32/FP64) is related to VRS. Most of the articles I've read explaining VRS in layman's terms basically use an example of rendering at lower resolutions, 2x1 or 1x4 pixel chunks for example. But what if there's a simple "Enable VRS" option that calculates an objects size and velocity on screen and lowers it's shading accuracy to match.

VRS is tweak to MSAA.

These videos show the basic idea on how it can be used.

INT/FP formats are part of ALU.

Yep, you guys are right, INT only for 4 and 8 bit. I was wondering how bad 4bit shaders might look. I used to own a GeforceFX after all.

Int4 has values from 0 to 15.
Int8 0-255.

I'm sure old 8bit or Amiga/AtariST coders will think something they could be used for.

 

[goonergaz]

@MrFox's post got me thinking.

Xsx's PSU is 315 watt, which indicates Xsx's TDP to be around 235W. This almost guarantees that the console is the highest TDP console in history, easily 35W more than the launch PS3s.

If you look at the TDP's that the highest end GPU's draw, they're usually 300-350w.

For a MAX console someday, we'll see a 400-500mm2 SOC, 300-350 watt TDP, 400-500 watt PSU. One can dream.

beaten to it by @MrFox, but also I did mention back in the initial reveal when everyone was saying about how XSX would draw serious power that it only had a figure 8 power lead vs launch PS3 and launch PS4pro so likely would not be the most power hungry console.

 

[Kaotik]

What for? What are the reasons to saturate CPU?

Every single thing in a game aside rendering, and even part of that goes to CPU?

 

[Dictator]

So you're suggesting that PS5 is going to render everything in greater than 4K native and scale down to 4K? Just to take advantage of their hard drive throughput?

And suffer from horrible aliasing. Textures use mip maps of course so I am not sure why someone would think every texture on screen is at mip 0 all the time. It would Look horrible and be wasteful.

 

[Shifty Geezer]

Speculation time - what cooling tech could Sony be employing that we'll like the look of as per Cerny's expectations, and can deal with the predict thermals? Is there anything new? Will their double-sided cooling come into play? Will the board be mounted centrally in the case with space either side, and a fan draw air over both front and back sides? And new magical techs?

 

[psorcerer]

are you referring to a texture being 4096x4096 as a 4K texture and a 8K texture as 8192x8192?
Because I'm not referring to that. Though in retrospect I should have.

I was thinking that if you stood the closest you could to a texture with the camera and the resolution of that texture still maintained 1:1 at native resolution with no stretching, then there would be no need to go higher. i don't actually know what texture size that needs to be for that to happen though.

mind you, as texture sizes get higher in size it is murderous on bandwidth with Ansio.

You're suggesting PS5 run 8K and 16K textures respectively?
8192x8192 and 16384x16384?

As higher rez mips are needed only for angled textures, and only for a small part at that, here things like "sampler feedback" would be nice to implement. Which means although you can use 16K texture, only a small part of it will be loaded into RAM at any given time.
Although I would like to see real texture space shading here (people told me that what is called TSS in modern NV and AMD is in fact just a fancy name for MSAA samples access).

 

[fehu]

Something that from the various video I've not understood: does the metal plate under the xsx's soc double as a mild heatsink similar to sony's patent?

 

[psorcerer]

MS beat him in hardware design this time around

Are you sure you know what word "hardware" means?
Because "VRS" and "Mesh Shaders" are both DirectX software APIs.
What part of them is really implemented in hardware and how it's called in hardware we don't know yet.

Every single thing in a game aside rendering, and even part of that goes to CPU?

That's an interesting solution. So, you have a fast as hell GPU for math operations and you still use CPU for everything?
I hope nextgen developers are not that bad at hardware utilization. Probably Cerny also hopes so...

 

[Kaotik]

That's an interesting solution. So, you have a fast as hell GPU for math operations and you still use CPU for everything?
I hope nextgen developers are not that bad at hardware utilization. Probably Cerny also hopes so...

I oversimplified it, of course, there's plenty of compute happening on the GPU-side too, but it's usually rendering related one way or another. All the game logic however to my understanding does run on CPU and lot of the code can't be effectively be parallized which would be step 1 requirement for running it on GPU, not to mention going back and forth over PCIe-bus is relatively slow, so it would make little sense to throw things at GPU just because you could , when CPU even if it's slightly slower executing it would in the end run it all faster

 

[Kaotik]

Something that from the various video I've not understood: does the metal plate under the xsx's soc double as a mild heatsink similar to sony's patent?

Don't know about Sonys patent, but yes, that metal plate between the motherboards is meant to work as a heatsink, too.

 

[psorcerer]

All the game logic however to my understanding does run on CPU and lot of the code can't be effectively be parallized

We know from Doom Eternal that it can.
It's just that most of the developers don't bother.
But anyway, game logic will not draw much power, at all.

going back and forth over PCIe-bus is relatively slow

On PC. On console you can write and read to the same coherent RAM.
But I think I know where you were getting at: multiplatform games on PC indeed cannot use HSA RAM correctly, because the CPU->GPU sync will eat up all the dividends.
But I hope this gen PC won't drag consoles down too much, like it was last gen.

 

[BRiT]

Sony's statements would be inclusive of profiling data for all games across its platform, all 4000+.

Trouble with profiling your existing 4000+ games is they are all made to run on Jaguar CPUs so they tell you nothing about what NextGen CPU loads you will have. You would need to profile PC games to get a better idea of workloads.

 

[dskneo]

I seriously question Cerny honesty in his claims. While they are true, they are only honest if you view them from a manufacturer POV (costs) but dishonest when seen from the gpu/cpu numbers game.


To elaborate, what use is there to reach 3.5/2.23ghz if that means developers will have to cap their code to obey the power draw envelope. AKA reduce work per cycle!



Suddenly, performance is determined by developers regulating their code, instead of letting the GPU regulating its power draw/temps curve. That is how it became "deterministic" and "predictable". "MHZ" numbers became meaningless for comparative against series X.



Whats the win/win for Sony:

1 - Frequency numbers on PS5 became superfluous because its no longer frequency that determines performance, but still allows them to claim 10TF and appear to the general public like they are still in the ball park.

2 - Cut BOM down (or better distribute the BOM to what most needs it) because you no longer need to overestimate future workloads.

 

[anexanhume]

Sony's statements would be inclusive of profiling data for all games across its platform, all 4000+. And out of that 4000+, there may be relatively few "Doom Eternal" type games from a utilization perspective since we don't have those statistics. But the "rarity of downclocking" statement could align with the significantly greater numbers of Indie and low budget games released on the platform versus actual usage or playtime in games likely impacted by downclocking, which consists of more optimized, higher load, AAA titles.

I doubt they’re profiling those games on that hardware for that purpose, and as such, the idea they’re skewing results with indie title performance is misleading and perhaps disingenuous. They’ve admitted they have only looked at ~100 games for boost mode, after all. They also have any internal titles they’d be working on.

Something that from the various video I've not understood: does the metal plate under the xsx's soc double as a mild heatsink similar to sony's patent?

It’s more a of structural and spacing element. It likely doesn’t have direct contact with any high dissipater, and its only thermal path is to boards with higher average temperatures. Thus, it will only provide thermal relief from airflow, for which it’s not shaped well.