No DX12 Software is Suitable for Benchmarking *spawn*

I wasn't sure I figured someone here would know. I quickly checked the Nvidia product page and came up empty handed.

 

PCGH benched Grid:Legens: https://www.pcgameshardware.de/GRID...nds-Benchmarks-Technical-Review-1389979/2/#a2

Another broken DX12 game on nVidia hardware - less than 110 FPS in 1080p.

 

UL preparing 3DMark Speed Way benchmark focusing on DirectX 12 Ultimate features

 

Meh :/ I am pretty disappointed, this doesn't look impressive to me. Especially considering its using mesh shading.

 

Sorry I dont read german (despite taking it in school) how is it broken?

 

It isn't, it's just the usual mantra. If AMD isn't doing as well as expected, it's AMDs fault, if NVIDIA isn't doing as well as expected, it's games fault.
If troyan had checked anything but graphs he'd know it too, as the review explicitly states it: Ego engine is just well optimized for AMD hardware, just like it has always been.

 

Nvidia is slower than AMD.

 

Your mantra? Don't remember anyone else here saying this.

Which means that it is badly optimized for Nvidia h/w because otherwise the results would be as in titles which are "just well optimized" for all h/w.

 

You should explain why outdated games like Grid:Legens or Valhalla can only archive ~100 FPS on nVidia hardware when at the same time i can play games with Raytracing at higher frames - Battlefield 5, Control, Shadow of the Tomb Raider, Formula 1 2021 etc.

DX12 has it own "DX11" moment: Software bottleneck limiting modern GPUs.

 

I too had german in school and I couldn't find any hint for this either. Just the usual mention of Codemasters games running above average on Radeon compared to GeForce.

 

Developers don't really have to use mesh shading solely for rendering compressed geometry. You can use mesh shading to do a straight replacement of the vertex shading pipeline in the trivial case. Also you could use lower resolution LODs for ray tracing just like how UE5 Lumen does for Nanite meshes ...

 

GPUScore Relic Of Life benchmarks with 22 GPUs - Article Page 3 - D3D12, OpenGL 4.5 and Vulkan (guru3d.com)

 

Something isn't right there. Look at the gap between the 3060 and 3060ti.

 

March 22, 2022

 

It's always a bit amusing seeing GPU reviews. You'd almost think that absolutely no titles in existence are optimized for NVidia GPU hardware and only AMD hardware receives any optimizations. I don't believe I heard the author of that video mentioning even once that NV might be performing better because a title might be optimized for NV hardware. But man, if the AMD GPU had a significant lead, it's almost always because the title must have been optimized for AMD hardware.

Note: I'm not saying a game might not be more optimized for AMD hardware, but I think it's a bit naive at best to believe that no titles receive any optimizations for NV hardware.

I mean it should be obvious that both vendor's GPUs receive optimizations in games to a greater or lesser degree with NV hardware generally benefitting from more optimizations just due to NV spending more than AMD on devrel.

Regards,
SB

 

Absolutely, I get that. But this sort of reinforces the weird stereotype that the only way AMD GPUs can win benchmarks is if they are optimized for AMD GPUs, while at the same time not acknowledging that NV are generally winning benchmarks because of optimizations for NV GPUs.

I mean, it's almost like optimizations for one GPU hardware arch. will give it an advantage in benchmarks. But the only time it's ever brought up is if AMD is benefiting, which sort of gives the impression that AMD's hardware is somehow inherently worse when in fact it's more of a case of which hardware arch any given title is receiving the most optimizations. RT aside, of course, because the RT on AMD hardware is so far behind that there just isn't much of a comparison there.

I don't even think most reviewers are consciously doing this (that would be unprofessional and borderline malicious), but it does go a long way towards promoting, IMO, bad information.

Better would be to just come out and say, hey. NV GPUs are typically a better buy because more titles will be optimized for NV GPUs than for AMD GPUs. Rather than only saying, well NV GPUs are better performers and the only time AMD GPUs can perform well is if a title is optimized for AMD GPUs. The first statement makes clear the current situation of gaming software relative to gaming hardware. The second implies that one IHVs hardware doesn't require optimizations to perform well in games while the other does, even if that wasn't the reviewer's intent.

Regards,
SB

 

It has been more and more intresting on the laptop side of things. Nice review.

 

DX12 continues to be slower than DX11 in the new game Tiny Tina Wonderlands which uses UE4, AMD cards are 20% faster with DX11, while NVIDIA cards are 10% faster.

https://www.computerbase.de/2022-03/tiny-tinas-wonderlands-benchmark-test/2/