No DX12 Software is Suitable for Benchmarking *spawn*

[BRiT]

Oh, its an update for Serious Sam Fusion... I thought it was an update for Steam itself since no context was given.

 

[Malo]

Oh, its an update for Serious Sam Fusion... I thought it was an update for Steam itself since no context was given.

Yay context. The idea that Steam itself was rendering something or affecting Vulkan/DX12 performance was a bit

 

[dogen]

Link corrected, they claim to the update improve performance under Direct3D 12 and Vulkan rendering mode. Surfing in the steam forum those rendering mode were not so optimized, especially the Vulkan mode which never gave better performance then Direct3D 11 rendering mode.

Seems a bit faster than D3D11 now. At least for my setup.

 

[Anarchist4000]

Yay context. The idea that Steam itself was rendering something or affecting Vulkan/DX12 performance was a bit

Steam has the overlay and some in game components that could cause problems. Especially for resource management, but yeah rather basic unless capping framerate to refresh or something.

 

[CarstenS]

DSOGaming headlines an editorial:
„The state of DX12 games in 2017, is DirectX 12 losing its steam?“

http://www.dsogaming.com/editorial/the-state-of-dx12-games-in-2017-is-directx-12-losing-its-steam/

 

[Anarchist4000]

Seems a bit disingenuous to exclude the exclusive DX12 titles when that's the only way to properly implement the models. API by all appearances works well, just Win10 adoption is a bit lackluster. Vulkan however has been rather impressive.

 

[CSI PC]

Seems a bit disingenuous to exclude the exclusive DX12 titles when that's the only way to properly implement the models. API by all appearances works well, just Win10 adoption is a bit lackluster. Vulkan however has been rather impressive.

Well I can relate why they are omitted.
Quantum Break was a DX12 exclusive and then done later to support DX11, where ironically showed decent optimisation.
In fact fair to say the DX12 experience for Nvidia was pretty bad, resolution probably down to the Nvidia engineers sorting out some of this in DX11 and their drivers.

On the flip side, surely there are some good exclusive DX12 titles that are clearly optimised-designed well for both AMD/Nvidia, but without DX11 it is difficult to know just how well.
I wonder where that leaves Nvidia with their DX12 Gameworks/libraries, will that end up being either an incentive or another issue in eyes of devs-studios.

 

[DavidGraham]

New Seems a bit disingenuous to exclude the exclusive DX12 titles when that's the only way to properly implement the models. API by all appearances works well, just Win10 adoption is a bit lackluster. Vulkan however has been rather impressive.

The article didn't exclude any DX12 exclusive title, they count them in like any other.

 

[CSI PC]

Did Nvidia ever explain or anyone found out what they did to gain up to 20% performance in Forza 7 with DX12 and a driver patch (387.92)?
Just curious if it was an issue Nvidia side or like both manufacturers do with DX11 overcome possible issues/inefficencies with the game code affecting them.

The only public announcement I found associated with Forza 7 and Nvidia driver patch update was:

The work doesn't end at launch, however, as our team continually searches for further improvements in game code and our drivers, collaborating with developers whenever possible.

Goes to show though how performance can be skewed just looking at launch, still relevant for consumers but as benchmarks something that needs to be monitored when talking about it a new game or actual benchmark.

 

[DavidGraham]

Warhammer Verminitide 2 just got released with both a DX11 and a DX12 path. As per usual, the DX12 gives worse avg and min fps on literally every GPU. Users are even reporting massive stutters and hitches only on DX12 even with AMD GPUs.

http://gamegpu.com/action-/-fps-/-tps/warhammer-vermintide-2-test-gpu-cpu

 

[Malo]

Why do they bother? Marketing checkbox?

 

[DavidGraham]

New Why do they bother? Marketing checkbox?

Yup, most developers just write "work in progress" under the toggle of DX12 and be done with it. Even Verminitide 2 is no exception to this. It's like this is their scapegoat, even 2.5 years after DX12 release!

 

[Rodéric]

D3D12 is hard and time consuming, but it's a marketing requirement.

 

[Malo]

Vulkaaaannnnnnn

 

[Deleted member 13524]

D3D12 is hard and time consuming, but it's a marketing requirement.

That's curious. Please, do expand.

Is this something that major dev teams agreed upon during conferences? Is it being pushed by publishers?
Are games with a DX12 label selling better on the PC?

New Vulkaaaannnnnnn

They certainly seem to be confident on overtaking DX12.

 

[Svensk Viking]

Warhammer Verminitide 2 just got released with both a DX11 and a DX12 path. As per usual, the DX12 gives worse avg and min fps on literally every GPU. Users are even reporting massive stutters and hitches only on DX12 even with AMD GPUs.

http://gamegpu.com/action-/-fps-/-tps/warhammer-vermintide-2-test-gpu-cpu

PCgameshardware.de published their results now. They say DX11 is better when GPU limited, and DX12 is better when CPU limited. They note that DX12 generally is better as soon as there are more enemies on screen.
http://www.pcgameshardware.de/Warha...s/Direct-X-12-Benchmarks-Test-Review-1252228/

 

[MfA]

We just don't have enough Carmacks in this world ... too many incompetents trying to jury rig decade old codepaths into DX12/Vulkan, piss poorly.

 

[DavidGraham]

PCgameshardware.de published their results now. They say DX11 is better when GPU limited, and DX12 is better when CPU limited. They note that DX12 generally is better as soon as there are more enemies on screen.

Actually they state fps under DX12 is generally lower than DX11, especially min fps. However DX11 seems to be limited @1080p for some reason while DX12 is not. Which is kind of a moot point when you are testing the game with high end GPUs that will operate @1440p or higher.

 

[CarstenS]

The WHVT-results are a bit inconklusive, I think. While the 1080 Ti and Vega56/64 gain even in low resolutions from using DX12, the GTX 1070 appears a bit slower, also the RX 580, even though it's only a tiny bit here.

 

[Anarchist4000]

Has NVIDIA improved DX12 and Async Compute performance with the Volta architecture? These were hot topics during Maxwell and Pascal’s reign. The general understood consensus is that AMD Polaris and Vega GPUs better handle DX12 and Async Compute performance, are more efficient at it and show a performance improvement advantage.

Since we had some time with the NVIDIA TITAN V we wanted to explore if DX12 and Async Compute efficiency and performance has improved with Volta versus Pascal. Now, this is hard to do apples-to-apples because the TITAN V is so much faster than the GeForce GTX 1080 Ti. Therefore, we cannot compare each video card’s gaming performance to each other. Instead what we must do is compare the difference between DX11 and DX12 performance on TITAN V to the difference between DX11 and DX12 performance on GTX 1080 Ti. This is not a "perfect" way to go about this, but we thought that the comparison still might have some merit and watned to see exactly what the data had to say. Given that TITAN V is the only card right now that gives us access to the VOLTA architecture, the comparison to the gains on Pascal still might at the very least be of interest to some of our readers.

This shows that if NVIDIA takes the Volta architecture and optimizes it and the drivers for gaming, we could see an uplift in DX12/Async Compute gaming. There appears to be a lot of headroom to improve upon. Already TITAN V, without optimized drivers, has shown these improvements to DX12 gaming over Pascal, so imagine what could happen with properly optimized drivers. We may finally see DX12 and Async Compute performance being advantageous with the new architecture.

NVIDIA VOLTA's TITAN V DX12 Performance Efficiency