guru3dDoom Vulkan Benchmarks

[dogen1]

I thought the point was to take stuff traditionally done on the CPU and do it on the GPU shaders..

Compute shaders can be used for GPGPU, but most of the time they're used for speeding up rendering.

 

[Bacon1]

I'll stick with OGL. That does allow for SLI.

I didn't think Doom supported SLI/CFX

 

[Arachnotronic]

I didn't think Doom supported SLI/CFX

It doesn't

 

[Racan]

Doom Vulkan R9 Fury X 4K 60fps Gameplay! Plus Fury X vs GTX 1070 Frame-Rate Tests!


DigitalFoundry finds a 26% advantage for the Fury X.

 

[Azix]

Doom Vulkan R9 Fury X 4K 60fps Gameplay! Plus Fury X vs GTX 1070 Frame-Rate Tests!


DigitalFoundry finds a 26% advantage for the Fury X.

AMD is in such a weird place. A year or two from now it could be the norm with dx12 and vulkan that the fury x is in this position. It could even be ahead of the 1080. But because we are at a transition point, the AMD cards have a split value. One they have now in dx11 games and some dx12 games, and one they might have later on. AMD has to sell the cards less than they could be worth based on dx11 performance.

probably said it before but 480 for example is good value in dx11, really good value in dx12/vulkan. but its judged by dx11

 

[RussianSensation]

They havent worked with nVidia. id just sold out. What a great gift to the nVidia customers who have bought the game in the last two months. :thumbsdown:

Doom is my last game from id. :thumbsup:

These statements of id selling out have proven wrong by many posters who provided quotes from Bethesda that the developers worked with both vendors. If id sold out to AMD, AMD would have never allowed them to release Doom with OpenGL which showed horrible performance on AMD cards prior to the post launch patches. Similarly, why would NV showcase Vulkan benches during their GTX1080 launch?

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

You should write a letter to NV why for the 3rd consecutive generation/architecture (Kepler, Maxwell, Pascal) they failed to include hardware Async Compute into their hardware given that DX12/Vulkan are the future of PC gaming and DX11/OpenGL are outdated API that no longer align with the direction modern GPU architectures are heading.

"So what is Vulkan exactly? Well, think of it as the OpenGL equivalent to DirectX 12, with many of the same advantages - principally, far better utilisation of multi-core CPUs, along with the implementation of GPU asynchronous compute. The latter element in particular sees big improvements for Radeon hardware, and it's used extensively in Doom. id Software's lead rendering programmer Tiago Sousa recently revealed efficiency improvements of 3-5ms per frame on the console versions of the game - a seriously big deal when you have a 16ms per-frame render budget. Vulkan allows us to finally code much more to the 'metal'. The thick driver layer is eliminated with Vulkan, which will give significant performance improvements that were not achievable on OpenGL or DX[11]."
http://www.eurogamer.net/articles/d...n-patch-shows-game-changing-performance-gains

Sure, you can use lame excuses how AMD's had inferior OpenGL drivers (although this isn't conclusive as in many OpenGL games 5870->7970Ghz outperformed GTX480->770), but the fact remains that newer DX12/Vulkan APIs allow better CPU multi-threading and higher GPU unit utilization for modern GPU architectures that can take advantage of it.

"In a tech interview with Digital Foundry (due to be published in full this weekend), the id team talk about the advantages of Vulkan and the potential of async compute in particular.

"Yes, async compute will be extensively used on the PC Vulkan version running on AMD hardware," lead programmer Billy Khan tells us. "Vulkan allows us to finally code much more to the 'metal'. The thick driver layer is eliminated with Vulkan, which will give significant performance improvements that were not achievable on OpenGL or DX."

Senior engine programmer Jean Geffroy goes into depth on the profound advantages that async compute brings to the table.

"When looking at GPU performance, something that becomes quite obvious right away is that some rendering passes barely use compute units. Shadow map rendering, as an example, is typically bottlenecked by fixed pipeline processing (eg rasterisation) and memory bandwidth rather than raw compute performance. This means that when rendering your shadow maps, if nothing is running in parallel, you're effectively wasting a lot of GPU processing power.

"Even geometry passes with more intensive shading computations will potentially not be able to consistently max out the compute units for numerous reasons related to the internal graphics pipeline. Whenever this occurs, async compute shaders can leverage those unused compute units for other tasks. This is the approach we took with Doom. Our post-processing and tone-mapping, for instance, run in parallel with a significant part of the graphics work. This is a good example of a situation where just scheduling your work differently across the graphics and compute queues can result in multi-ms gains.

"This is just one example, but generally speaking, async compute is a great tool to get the most out of the GPU. Whenever it is possible to overlap some memory-intensive work with some compute-intensive tasks, there's opportunity for performance gains. We use async compute just the same way on both consoles. There are some hardware differences when it comes to the number of available queues, but with the way we're scheduling our compute tasks, this actually wasn't all that important."

Seems you are just upset NV's Pascal hardware is once again missing the most important next generation advancement in improving multi-tasking for next gen GPU architectures - Hardware Async Compute. It's OK, NV will be just find this generation due to the raw horsepower and efficiency advantage of 16nm TSMC + Pascal architeture. Having said that, Pascal would have undoubtedly been better off had NV added hardware Async Compute, despite a die size increase and perf/watt penalties.

It's pretty telling that under a next gen API, the last gen's 28nm Fury X is beating GTX1070 by 26-27% at Computebase, and this how now been corroborated by Digital Foundry in a head-to-head video.

That would be akin to a GTX980Ti beating Vega by 25-30% under a DX12/Vulkan game.

Async Compute is only going to be used more and more.
http://www.eurogamer.net/articles/digitalfoundry-2016-doom-tech-interview

"Digital Foundry: Will we see async compute in the PC version via Vulkan?

Billy Khan: Yes, async compute will be extensively used on the PC Vulkan version running on AMD hardware. Vulkan allows us to finally code much more to the ;metal'. The thick driver layer is eliminated with Vulkan, which will give significant performance improvements that were not achievable on OpenGL or DX.

Digital Foundryo you foresee a time where async compute will be a major factor in all engines across formats?

Billy Khan: The time is now, really. Doom is already a clear example where async compute, when used properly, can make drastic enhancements to the performance and look of a game. Going forward, compute and async compute will be even more extensively used for idTech6. It is almost certain that more developers will take advantage of compute and async compute as they discover how to effectively use it in their games."

Still does not explain why a game like Doom is so much compute hungry to have a lineal perf/TFlops.

The technical interviews (see above) with the developers explain it well.

1. Vulkan API = closer to the metal => removes abstraction layers that bottlenecked certain GPUs
2. Maximizing Compute and Async Compute => increased utilization of GPU resources and CPU multi-threading.

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.

Pretty sad that even GameGPU dropped the ball on that one, instead choosing to test with SMAA. As far as the Polish sites go, it's not a surprise, esp. the insanely biased GPU reviews coming out of NVLabs.

"Digital Foundry: Can you talk us through how the 8x TSSAA implementation works? Is it consistent between consoles and PC?

Tiago Sousa: I've always been a fan of amortising/decoupling frame costs. TSSAA is essentially doing that - it reconstructs an approximately 8x super-sampled image from data acquired over several frames, via a mix of image reprojection and couple heuristics for the accumulation buffer.

Digital Foundry: Can you discuss the directional occlusion settings on PC?

Tiago Sousa: Lower settings use lower sample count, higher settings use higher sample count. We actually use a fairly low amount of samples overall, but rely on the TSSAA to reconstruct a higher quality result over frames. It's quite performant, about 0.1ms on PC at 1440p."

Per the DF video comparing GTX1070 vs. Fury X under Vulkan, a single Fury X was able to manage almost 50 fps average at 4K. The only other FPS that I can think of was this well optimized was Star Wars Battlefront. At first I feared Doom became a GW title after JHH showed it off during the GP104 unveil but it's good to see this didn't come true.

I would venture to guess that Volta would be a much better GPU architecture if it takes all the efficiencies of Pascal and adds more potent hardware compute/Async Compute functionality.

 

[HurleyBird]

A year or two from now it could be the norm with dx12 and vulkan that the fury x is in this position. It could even be ahead of the 1080.

If FuryX had an 8GB frame buffer like the 1080 does, then sure, that might be possible. The 4GB frame buffer will almost certainly become a limiting factor in the next couple of years though.

 

[AnandThenMan]

id Tech 6 is the best thing to happen to PC gaming in a very long time. Not tied to an OS/D3D is :thumbsup: :thumbsup: :thumbsup: :thumbsup:

 

[Bacon1]

If FuryX had an 8GB frame buffer like the 1080 does, then sure, that might be possible. The 4GB frame buffer will almost certainly become a limiting factor in the next couple of years though.

Will it though? As shown by digital foundry and others, Fury is keeping up and surpassing the 8gb 1070 @ 4k even with TSSAA 8x and Ultra settings. I think by the time the 4gb is hampering it, you'll be turning down settings anyway due to the processing power not being enough.

Devs have gone nuts with lacking compression, I mean look at the install of some games these days, 60+GB. Even games like shadow of mordor which say they require 5+gb for the ultra textures don't, and there isn't anything noticable in them anyway. They are wasteful. Optimization is key to all of this and which is why devs that do Vulkan/DX12 correctly will really stand out from the rest with how well their game runs and how well it looks.

 

[AdamK47]

It doesn't

NVidia has had an SLI profile for Doom for quite a while now in their drivers.

I ran the game with SLI on (red bar) and with SLI off (blue bar). This was from the very start of the game to the point where you go outside.

 

[dzoni2k2]

NVidia has had an SLI profile for Doom for quite a while now in their drivers.

I ran the game with SLI on (red bar) and with SLI off (blue bar). This was from the very start of the game to the point where you go outside.

I would hardly call that working though. Your GPU utilization when running SLI is around 50%.

 

[Erenhardt]

NVidia has had an SLI profile for Doom for quite a while now in their drivers.

I ran the game with SLI on (red bar) and with SLI off (blue bar). This was from the very start of the game to the point where you go outside.

Your graph suggest you would be far better of without SLI.

 

[AdamK47]

There is a performance benefit. It's not great, but it does divide up usage across both GPUs. I'll take that over slightly less performance and one GPU at 90 to 100%. I mean, why wouldn't you?

 

[gorobei]

.......

Async Compute is only going to be used more and more.
http://www.eurogamer.net/articles/digitalfoundry-2016-doom-tech-interview

"Digital Foundry: Will we see async compute in the PC version via Vulkan?

Billy Khan: Yes, async compute will be extensively used on the PC Vulkan version running on AMD hardware. Vulkan allows us to finally code much more to the ;metal'. The thick driver layer is eliminated with Vulkan, which will give significant performance improvements that were not achievable on OpenGL or DX.

Digital Foundryo you foresee a time where async compute will be a major factor in all engines across formats?

Billy Khan: The time is now, really. Doom is already a clear example where async compute, when used properly, can make drastic enhancements to the performance and look of a game. Going forward, compute and async compute will be even more extensively used for idTech6. It is almost certain that more developers will take advantage of compute and async compute as they discover how to effectively use it in their games."



The technical interviews (see above) with the developers explain it well.

1. Vulkan API = closer to the metal => removes abstraction layers that bottlenecked certain GPUs
2. Maximizing Compute and Async Compute => increased utilization of GPU resources and CPU multi-threading.

........

id Tech 6 is the best thing to happen to PC gaming in a very long time. Not tied to an OS/D3D is :thumbsup: :thumbsup: :thumbsup: :thumbsup:

vulkan does a lot for multiplatform devs and expands the target users.

in a talk about vulkan, one dev was saying that it allowed them to discard all the work they did for backwards compatibility for dx12 dx11 dx9. since they could utilize the vukan code for all three, whereas before they had to do 3 different versions. this allows them to target older hardware and OS by just having one vulkan driver.

 

[Erenhardt]

AdamK47, Frame pacing and frame latency? At almost the same framerate, one GPU is superior to two, at least that was used against 2x480 by some

 

[AdamK47]

I don't feel a difference at all. Do you feel a difference in the game when you run SLI?

 

[Erenhardt]

Yes, I do.

 

[dzoni2k2]

There is a performance benefit. It's not great, but it does divide up usage across both GPUs. I'll take that over slightly less performance and one GPU at 90 to 100%. I mean, why wouldn't you?

I'm sure you get a small boost it's just that you likely don't need it in a game where you already get close to 200 FPS. You really don't get your moneys worth back.

 

[AdamK47]

Yes, I do.

Oh, very interesting. What cards are you running in SLI?

I'm sure you get a small boost it's just that you likely don't need it in a game where you already get close to 200 FPS. You really don't get your moneys worth back.

I'm not going to exclusively play Doom with my PC. Besides, "moneys worth" really doesn't apply to me.

 

[Red Hawk]

Doom Vulkan R9 Fury X 4K 60fps Gameplay! Plus Fury X vs GTX 1070 Frame-Rate Tests!


DigitalFoundry finds a 26% advantage for the Fury X.

"The big challenge I actually had with Doom on the R9 Fury X was to actually make it run slower than 60 FPS". Nice.

 

[Bacon1]

vulkan does a lot for multiplatform devs and expands the target users.

in a talk about vulkan, one dev was saying that it allowed them to discard all the work they did for backwards compatibility for dx12 dx11 dx9. since they could utilize the vukan code for all three, whereas before they had to do 3 different versions. this allows them to target older hardware and OS by just having one vulkan driver.

Yep, Vulkan = Win7->10, plus linux (if rest of game is coded to support it).

DX12 = Win10

Now the only downside is losing support for older hardware, but I think we'll start seeing that in the next year, which means we'll have either pure vulkan, pure dx12 and dx12+vulkan only games which remove opengl / DX11 support all together. Devs that want to support OS other than Win10 will include Vulkan over DX11, and assuming that DX12 is faster (or easier to code for for xb1) DX12 as well.

 

[atakall]

They havent worked with nVidia. id just sold out. What a great gift to the nVidia customers who have bought the game in the last two months. :thumbsdown:

Doom is my last game from id. :thumbsup:

You're incorrect. On May 6, 2016, nVidia held a conference to unveal the 1080. During that conference, 2 employees of iD were invited on-stage to discuss working with nVidia in connection with Doom and to showcase footage of a GTX1080 playing Doom. The video of the relevant segment can be viewed at https://www.youtube.com/watch?v=ueRjbYdcXbs. During the video, iD acknowledges that nVidia engineers worked closely with iD for months in connection with Doom, not only on OpenGL but also Vulkan. The extent of the cooperation included at least a two week period where nVidia engineers worked from iD offices.

The relevant quotes, along with the time that they appear in the video, are as follows:

At 0:09
id: "They've [nVidia] had engineers, actually, at our offices for the last week and all of next week too working with us on the OpenGL stuff for the launch and then Vulkan as well. . . . we've been optimizing OpenGL for the launch."

At 5:00
nVidia: "How's it been like working with nVidia, Pascal, Vulkan, how's it been."
id: "Really good. Again, we've had drivers on site off and on for the last couple months but pretty heavy for the last few weeks working through OpenGL issues . . . "

 

[dzoni2k2]

You're incorrect. On May 7, 2016, nVidia held a conference to unveal the 1080. During that conference, 2 employees of iD were invited on-stage to discuss working with nVidia in connection with Doom and to showcase footage of a GTX1080 playing Doom. The video of the relevant segment can be viewed at https://www.youtube.com/watch?v=ueRjbYdcXbs. During the video, iD acknowledges that nVidia engineers worked closely with iD for months in connection with Doom, not only on OpenGL but also Vulkan. The extent of the cooperation included at least a two week period where nVidia engineers worked from iD offices.

The relevant quotes, along with the time that they appear in the video, are as follows:

At 0:09
id: "They've [nVidia] had engineers, actually, at our offices for the last week and all of next week too working with us on the OpenGL stuff for the launch and then Vulkan as well. . . . we've been optimizing OpenGL for the launch."

At 5:00
nVidia: "How's it been like working with nVidia, Pascal, Vulkan, how's it been."
id: "Really good. Again, we've had drivers on site off and on for the last couple months but pretty heavy for the last few weeks working through OpenGL issues . . . "

Talk about getting rekt with hard facts

 

[Red Hawk]

Yeah, I highly doubt Nvidia would have showcased Doom running on a 1080 with Vulkan before the patch even publicly released if id wasn't working with Nvidia. And it would be really odd for id to "sell out" so quickly after Doom initially released with an Nvidia advantage in OpenGL.

I have a feeling sontin might be a bit disingenuous with his reaction, though.

 

[AdamK47]

Talk about getting rekt with hard facts

We need more FUD busting on this forum. There are far too many instances where people say something completely wrong and it's cycled through comments as if it were factual. Most of the time it comes from passed around anecdotal second hand information. I generally stick to my own research of the facts or simply experience things first hand. I will say though, there are usually some good nuggets of facts posted here. I generally get a second opinion when I do read them. That's the broader nature of the Internet though.