davis.anthony
Veteran
Digital Foundry Article Technical Discussion [2024]
- Thread starter Shifty Geezer
- Start date
snc
Veteran
I asked about example when lower amd gpu outperform higher model on same architecture but with lower clocks, I have never seen such a case edit: In apparently old engine cyberpunk 6800xt wider but lower clocks demolish 6700xt
davis.anthony
Veteran
snc
Veteran
Hehe thats only theory that tried to explain performance difference and I show you is not very good one as we can compare using pc gpus on same arch. with higher cores number and lower clocks that beat lower cores higher clocks (as surprise surpise tflops were good performance indicator on same arch.)
arandomguy
Veteran
You'd have to test this basically by manually overclocking/underclocking with desktop GPUs as it doesn't really make sense to actually configure desktop end products that way, certainly not the extent of the PS4/XSX, on the PC side. Even then it wouldn't capture the same scenario.
The 6700xt and 6800 comparisons don't work because the 6800XT still scales roughly proportionally higher in all aspects (ROPs and other units aside from shaders, memory bandwidth without caveats, proportionally higher power budget, etc.) with no drawbacks/limitations and is running with the same driver/api/os stack and is actually architecture identical. This isn't the case with the PS5 and XSX.
Thinking about the best example on the PC side that might highlight why it's not just theoretical TFLOPs and GPU "width" is looking at AMD's APUs in final products particularly with the Steam Deck also thrown in there. The Steam Deck's APUs is just much weaker on paper than the now newer alternatives and while generally slower it's not always the case and the gap is smaller than the TFLOPs numbers would often indicate. Even just the Z1 vs. Z1 Extreme numbers don't show anywhere near the gap the tflops indicates, and that is much larger proportionally than the PS5 vs XSX.
The 6700xt and 6800 comparisons don't work because the 6800XT still scales roughly proportionally higher in all aspects (ROPs and other units aside from shaders, memory bandwidth without caveats, proportionally higher power budget, etc.) with no drawbacks/limitations and is running with the same driver/api/os stack and is actually architecture identical. This isn't the case with the PS5 and XSX.
Thinking about the best example on the PC side that might highlight why it's not just theoretical TFLOPs and GPU "width" is looking at AMD's APUs in final products particularly with the Steam Deck also thrown in there. The Steam Deck's APUs is just much weaker on paper than the now newer alternatives and while generally slower it's not always the case and the gap is smaller than the TFLOPs numbers would often indicate. Even just the Z1 vs. Z1 Extreme numbers don't show anywhere near the gap the tflops indicates, and that is much larger proportionally than the PS5 vs XSX.
Metal Snake Eater
Newcomer
is it a Benchmark though?
Richard stated there is DRS going on and "Series S running with a reduced resolution is able to out-perform Series X" does it mean Series S is a better console?
and "perhaps the dynamic resolution formula is different between consoles and PC"
more academic research using the PC benchmark pass than Actual Benchmark pass
if we rely on Dynamic Res Scaling we'll get results (PS5/XSX delivering 3080 performance again .. 3090 next time it's only 10-15% higher than 3080 afterall) so we're getting there
Richard stated there is DRS going on and "Series S running with a reduced resolution is able to out-perform Series X" does it mean Series S is a better console?
and "perhaps the dynamic resolution formula is different between consoles and PC"
more academic research using the PC benchmark pass than Actual Benchmark pass
if we rely on Dynamic Res Scaling we'll get results (PS5/XSX delivering 3080 performance again .. 3090 next time it's only 10-15% higher than 3080 afterall) so we're getting there
davis.anthony
Veteran
Mesh shader support on PlayStation is very close to pc. I am shocked. SHOCKED.
I think the reality is the Xbox and PlayStation architectures are probably almost identical when it comes to cpu and gpu families.
I'd be willing to bet the sampler feedback logic is also inside PS5's GPU.
I'm not so sure. DX12U required certain things of RDNA2/Series consoles, Sony weren't forced to require such things.
Sony focused on a more limited set of features, and so far, they were not wrong to have done so. They've already won the generation.
Why would you bet that? If it was, Sony almost assuredly would have said so since MS was making a big point of using it as a selling point.
There's been more than enough evidence by now to support the claim that the PS5 processor simply isn't properly DX12U compliant and was probably a big reason that they could get an early headstart for dev kits and mass production.
I don't think this is the reason they don't use all DX12 functionalities. They wanted an APU around 300 mm2 and they decided to choose the RDNA 2 functionalities they think useful for PS5 within this APU size. They choose this size because they were thinking about cost reduction during the lifecycle of the PS5.
EDIT: 308 mm2 to be precise.
davis.anthony
Veteran
As PS5 doesn't use DX, it's not anything DX compliant.
But have Sony ever said mesh shades work on PS5?
We said they didn’t support amplification shaders.
The thought process was that if PS5 is missing per primitive output; it’s on the older RDNA1 geometry front end. The rdna2 support per primitive output on mesh shader. And latter also support amplification shaders which don’t exist on rdna1.
We’ve cited several times that only difference between Mesh Shader support on DX12U requires per primitive output. If PS5 is almost PC, then this would be accurate. I think technically people are correct. Nvidias standard won out on dx12 and not AMDs.
Current Generation Games Analysis Technical Discussion [2023] [XBSX|S, PS5, PC]
The end result though amounted from pretty satisfactory to just perfect. Its been so long that I cant remember if they ever managed to replicate a transparent reflection using this technique ever (like seing your reflection on a window and at the same time whats behind it). I am checking MGS2's...
forum.beyond3d.com
Last edited:
Similar threads
