guru3dDoom Vulkan Benchmarks

[guachi]

I think the real take away here is that DOOM is badass. I'll say it again. DOOM is that ass while being the opposite of good, in an awesome way.

The performance increases are a great marketing coup for both AMD and Doom. Gets them both in the news for positive, real improvements. And those improvements came while the game is still new (so new sales = more $)

Because then people say stuff like the above (and I assume you really mean it).

I haven't played Doom since... umm.... Doom 1. I'm pretty sure I played the shareware version.

 

[sweetusernames]

Is the touted "GCN effect" finally coming into play?

 

[brandonmatic]

Interesting to see how Vulkan Doom performance correlates closely to theoretical peak compute:

• 970 -- 3.4 TFlops vs. 60.6 FPS
• 390 -- 5.1 TFlops vs. 77.4 FPS
• 480 -- 5.2 TFlops vs. 79 FPS
• 980ti -- 6.1 TFlops vs. 85 FPS
• 1070 -- 5.8 TFlops vs. 88 FPS
• Fury X -- 8.6 TFlops vs. 110 FPS

It's not perfect correlation, but it seems pretty telling.

https://www.computerbase.de/2016-07...md-nvidia/#diagramm-doom-mit-vulkan-2560-1440
http://www.pcper.com/reviews/Graphi...-RX-480-will-launch-199-more-5-TFLOPS-compute
http://techreport.com/review/28513/amd-radeon-r9-fury-x-graphics-card-reviewed/4

 

[Paratus]

Interesting to see how Vulkan Doom performance correlates closely to theoretical peak compute:

• 970 -- 3.4 TFlops vs. 60.6 FPS
• 390 -- 5.1 TFlops vs. 77.4 FPS
• 480 -- 5.2 TFlops vs. 79 FPS
• 980ti -- 6.1 TFlops vs. 85 FPS
• 1070 -- 5.8 TFlops vs. 88 FPS
• Fury X -- 8.6 TFlops vs. 110 FPS

It's not perfect correlation, but it seems pretty telling.

https://www.computerbase.de/2016-07...md-nvidia/#diagramm-doom-mit-vulkan-2560-1440
http://www.pcper.com/reviews/Graphi...-RX-480-will-launch-199-more-5-TFLOPS-compute
http://techreport.com/review/28513/amd-radeon-r9-fury-x-graphics-card-reviewed/4

So I'm rocking an OC'd 7970ghz at 4+ TFlops. Are you saying it might pull better frames than a 970.

 

[brandonmatic]

So I'm rocking an OC'd 7970ghz at 4+ TFlops. Are you saying it might pull better frames than a 970.

Ha, I don't know. That's a good test of the theory!

 

[boozzer]

Interesting to see how Vulkan Doom performance correlates closely to theoretical peak compute:

• 970 -- 3.4 TFlops vs. 60.6 FPS
• 390 -- 5.1 TFlops vs. 77.4 FPS
• 480 -- 5.2 TFlops vs. 79 FPS
• 980ti -- 6.1 TFlops vs. 85 FPS
• 1070 -- 5.8 TFlops vs. 88 FPS
• Fury X -- 8.6 TFlops vs. 110 FPS

It's not perfect correlation, but it seems pretty telling.

https://www.computerbase.de/2016-07...md-nvidia/#diagramm-doom-mit-vulkan-2560-1440
http://www.pcper.com/reviews/Graphi...-RX-480-will-launch-199-more-5-TFLOPS-compute
http://techreport.com/review/28513/amd-radeon-r9-fury-x-graphics-card-reviewed/4

that is as good as solid evidence. GG amd.

 

[Bacon1]

Interesting to see how Vulkan Doom performance correlates closely to theoretical peak compute:

• 970 -- 3.4 TFlops vs. 60.6 FPS
• 390 -- 5.1 TFlops vs. 77.4 FPS
• 480 -- 5.2 TFlops vs. 79 FPS
• 980ti -- 6.1 TFlops vs. 85 FPS
• 1070 -- 5.8 TFlops vs. 88 FPS
• Fury X -- 8.6 TFlops vs. 110 FPS

It's not perfect correlation, but it seems pretty telling.

https://www.computerbase.de/2016-07...md-nvidia/#diagramm-doom-mit-vulkan-2560-1440
http://www.pcper.com/reviews/Graphi...-RX-480-will-launch-199-more-5-TFLOPS-compute
http://techreport.com/review/28513/amd-radeon-r9-fury-x-graphics-card-reviewed/4

To compare the FPS vs TFlops directly:

AMD

390: 77.4 / 5.1 = 15.17 fps/flop
480: 79 / 5.2 = 15.19 fps/flop
Fury X: 110 / 8.6 = 12.79 fps/flop (cpu bound still or something else???)

Nvidia

970: 60.6 / 3.4 = 17.82 fps/flop
980 TI: 85 / 6.1 = 13.93 fps/flop
1070: 88 / 5.8 = 15.17 fps/flop

 

[Riek]

To compare the FPS vs TFlops directly:

AMD

390: 77.4 / 5.1 = 15.17 fps/flop
480: 79 / 5.2 = 15.19 fps/flop
Fury X: 110 / 8.6 = 12.79 fps/flop (cpu bound still or something else???)

Nvidia

970: 60.6 / 3.4 = 17.82 fps/flop
980 TI: 85 / 6.1 = 13.93 fps/flop
1070: 88 / 5.8 = 15.17 fps/flop

Bit misleading as the original flop numbers are completely wrong.

1070 has a flop rating of 6.4TFLop. That is without taking the overboost into account.
Could range anyware from 6.4 to 7.2TFlop (1.9GHz)

Same story for the 970 980,..

980 ti has 6Tflops that can range up to 6.7TFlops (1.2GHz)

Also the numbers of AMD 480 is misleading as it gives max 5.8TFlops


edit : 970 would push 3.9Tflops (not 3.4) probably ranging up to 4.2TFlops (1.25GHz)

 

[seitur]

So I'm rocking an OC'd 7970ghz at 4+ TFlops. Are you saying it might pull better frames than a 970.

Get Doom, test it and post results here

 

[Erenhardt]

Not every review shows a massive win for amd.
http://www.purepc.pl/karty_graficzne/doom_z_api_vulkan_wybuchowy_test_wydajnosci_kart_graficznych
Here, cards are right were you expect them to be.

Looks like they have nice 980ti sample, as it is faster than 1070. They don't pre-heat their GPUs so that could impact their results.
480 is at the nice spot too.

 

[Silverforce11]

Not every review shows a massive win for amd.
http://www.purepc.pl/karty_graficzne/doom_z_api_vulkan_wybuchowy_test_wydajnosci_kart_graficznych
Here, cards are right were you expect them to be.

Looks like they have nice 980ti sample, as it is faster than 1070. They don't pre-heat their GPUs so that could impact their results.
480 is at the nice spot too.

How lovely, those polish sites coming to the rescue...

SMAA. TSSAA was specially developed for Doom, it's used on the consoles to provide the best AA technique.

On the current Vulkan patch, you may or may not be aware, but id Software has confirmed, Async Compute is only active with no AA or TSSAA. They will add support for other AA modes soon.

 

[DDH]

How lovely, those polish sites coming to the rescue...

SMAA. TSSAA was specially developed for Doom, it's used on the consoles to provide the best AA technique.

On the current Vulkan patch, you may or may not be aware, but id Software has confirmed, Async Compute is only active with no AA or TSSAA. They will add support for other AA modes soon.

Not only that, that graph is just vulkan.

Heres open gl

FuryX gains 59%! Without async!



Ill say it. Every review shows a massive win for AMD

 

[Erenhardt]

How lovely, those polish sites coming to the rescue...

SMAA. TSSAA was specially developed for Doom, it's used on the consoles to provide the best AA technique.

On the current Vulkan patch, you may or may not be aware, but id Software has confirmed, Async Compute is only active with no AA or TSSAA. They will add support for other AA modes soon.

Interesting, why would they use MSSA for 1440p and TSSAA for other tests? Maybe error on the graph?

 

[seitur]

Interesting, why would they use MSSA for 1440p and TSSAA for other tests? Maybe error on the graph?

There is no AA for 4k, so diffrent AA for diffrent resolutions seem to be on purpose.

 

[amenx]

Question is.. is it a one-off? Did Nvidia miss something in driver optimization (ie, not address Vulkan performance in current drivers)? Because in another Vulkan based game, Dota 2, NV sees appreciable gains vs DX11:

https://www.computerbase.de/2016-05...gramm-dota-2-reborn-1920-1080-i7-6700k-replay

No other Vulkan based games other than AOS and Talos Principle (latter with poor results for both AMD/NV).

 

[3DVagabond]

NV boost probably coming soon:

Waiting for Async support in nVidia drivers? We've heard that line before and we're still waiting.

 

[Erenhardt]

Question is.. is it a one-off? Did Nvidia miss something in driver optimization (ie, not address Vulkan performance in current drivers)? Because in another Vulkan based game, Dota 2, NV sees appreciable gains vs DX11:

https://www.computerbase.de/2016-05...gramm-dota-2-reborn-1920-1080-i7-6700k-replay

No other Vulkan based games other than AOS and Talos Principle (latter with poor results for both AMD/NV).

There is dota2, but it looks like Doom (pun) for nvidia (degrades perf for nv improves for amd)
http://www.phoronix.com/scan.php?page=article&item=amdgpu-rx480-linux&num=11

 

[3DVagabond]

Apparently happened a while ago:

http://www.gamespot.com/forums/pc-a...ly-enables-async-with-march-drivers-33047280/

IIRC, we have seem some boost for Maxwell in DX12 titles.

I know what the title of that thread says but I can't see in the posts where it's happened.

 

[kondziowy]

Interesting, why would they use MSSA for 1440p and TSSAA for other tests? Maybe error on the graph?

It's a weird decision. In area I tested on Fury 1440p with SMAA you get 80-90fps, and with 8xTSSAA you get 95-105 fps. TSSAA also removes ugly pixels from water, and removes the moving jagged edges.
So with SMAA you get worse performance and worse image quality.

 

[amenx]

There is dota2, but it looks like Doom (pun) for nvidia (degrades perf for nv improves for amd)
http://www.phoronix.com/scan.php?page=article&item=amdgpu-rx480-linux&num=11

Yep.... on Linux.

 

[Erenhardt]

Yep.... on Linux.

You blame drivers?

 

[3DVagabond]

DX12 was never promised for Rise of the Tomb Raider. Nixxess has not published a FAQ where they blamed AMD for the "bad" performance. They havent tweeted hundreds of tweets to show how great nVidia is.

I promise you you wont see any improvements for nVidia from id. Not one. Mark my words.

It's not like they can redesign nVidia's hardware or rewrite the drivers for them.

 

[Erenhardt]

Can you run Doom on Vulkan under linux?

 

[guskline]

SilverForce11, is CF enabled in DOOM yet? I own DOOM but both of my rigs have single cards. Since you are running 2 RX480s in CF could you tell me if your fps jumps?

Thanks.

 

[lixlax]

Interesting to see how Vulkan Doom performance correlates closely to theoretical peak compute:

• 970 -- 3.4 TFlops vs. 60.6 FPS
• 390 -- 5.1 TFlops vs. 77.4 FPS
• 480 -- 5.2 TFlops vs. 79 FPS
• 980ti -- 6.1 TFlops vs. 85 FPS
• 1070 -- 5.8 TFlops vs. 88 FPS
• Fury X -- 8.6 TFlops vs. 110 FPS

It's not perfect correlation, but it seems pretty telling.

https://www.computerbase.de/2016-07...md-nvidia/#diagramm-doom-mit-vulkan-2560-1440
http://www.pcper.com/reviews/Graphi...-RX-480-will-launch-199-more-5-TFLOPS-compute
http://techreport.com/review/28513/amd-radeon-r9-fury-x-graphics-card-reviewed/4

I think you did hit the nails head. It clearly shows how good hardware AMD actually has and also how good work Nvidia has been doing on their drivers. Even if you look at the linked Dota 2 benchmark we see that 780ti is right where it should be regarding the raw TFLOPS: faster than 970 and 980. And it probably would be there in DX11 titles as well if Kepler still had the driver support that Pascal and Maxwell? are enjoying today.
This bodes well for AMD if future games are considered. All of the sudden, RX480 doesn't look as bad of an investment as it did after reading the reviews.