The secret of Doom (2016) performance on AMD GPUs

[the54thvoid]

I wonder what will the fan boys respond when nvidia shift to ALU heavy structure in Volta. Welp they are probably gonna switch their stance immediately to praise DX12/Vulkan.

In a fanboy's world if nvidia is doing bad it is AMD to blame some how.

You can't fathom the mind of one unless you are one, so why try?
I don't think Nvidia have any issues with their current implementation. They may have a shift back to compute and large parallelism to help with Asynchronous loads but it's a trade off from efficiency and clock speed.
People keep overlooking the 'limited' hardware under GP104's hood and neglect to acknowledge it (1080) beats all single GPU implementations by AMD GCN cards.
However, what I don't like about current Nvidia practice is the insane pricing. But this is due to no competition. 1060, price close to and often matching 480. 1070 cheaper than Fury X (which is a close match). 1080 and Titan X, no competition, price is stratospheric.

 

[Vayra86]

Heh, NVIDIA again "working" on driver support. Just like for async on GTX 900 series, right? :/

Dream on

 

[Ahhzz]

Can't believe how rabid people are in BOTH camps. Rarely does any conversation of "this runs better here or there" not turn into a complete fest of "Oh yeah?!! what about when your guys did this or that??!!" Just accept that both camps are doing their best to get an advantage over the other camp, and both of them have, at times, and will do, in the future, things that are "questionable" to get that edge. Welcome to healthy competition. Just accept that one camp or the other does better on this game or that, for now, and Let It Go!

 

[eidairaman1]

Sounds like a ticked greeny if you ask me. Performance gains are always welcome no matter Green or Red

 

[Fluffmeister]

As has been noted elsewhere it seems shiny modern Intel CPU's also help:

The game runs great on nV hardware regardless of all the drama.

 

[Xzibit]

Heh, NVIDIA again "working" on driver support. Just like for async on GTX 900 series, right? :/

You never know. Now that the 900 series is EOL they might get around to it.

Expect the same thing with Pascal Async Compute Preemption. Nvidia is going to sponsor a lot of titles and ports through their GameWorks programs to counter gains given PS4/PS4 Neo & XBOX1/XBOXScorpio are going to be developers major money makers like always 3:1 vs PC.

 

[GoldenX]

It's an indie emulator. Do you seriously believe they'll invest driver team time into it compared to games sold in millions of copies on a market that heavily depends on a functional ecosystem?...

They should, any bug reported for an API helps all users and cases.

The Khronos presentation (OP picture) also showed that developers are screaming for more stable Vulkan drivers, to both camps. They aparently get BSODs and broken extensions constantly.
And Nvidia got some flak for not giving OpenCL 2.X support

 

[arbiter]

Hair works uses standard D3D APIs - nothing NVIDIA specific. This argument is invalid. Now that AMD has reimplemented tesselation in Polaris and the Witcher runs fine even with HairWorks on on AMD everyone has magically forgotten about this fact.
And if I'm not mistaken W1zzard doesn't use PhysX enabled games in his testbed (or disables it completely). So PhysX argument is equally invalid.

Hair works used DX11 Tessellation which has been apart of DX11 since day 1. as for PhysX, that is set to off in any game that was used it so it wasn't impacting anything in testing.

Is vulcan a closed system like PhysX and hair works? What are the barriers for NVIDIA using this game?

Vulkan USED to be mantle which was AMD closed proprietary API, AMD said they would open it up to all but as everyone expect that never happened and it stayed close source til AMD dropped the project and gave source to people that make OpenGL. However vulkan is still optimized and coded for AMD cards specifically so they get an advantage from it even though its not open source AMD has a good 2-3 or so year head start with it.

I wonder what will the fan boys respond when nvidia shift to ALU heavy structure in Volta. Welp they are probably gonna switch their stance immediately to praise DX12/Vulkan.

In a fanboy's world if nvidia is doing bad it is AMD to blame some how.

AMD's drivers are still miles behind, they have way more cpu overhead it seems when you are more cpu limited. Was a site that tested some older intel and amd cpu and under vulkan, RX 480 took pretty sizeable hit
Give size of image i will just provide a link for it, http://i.imgur.com/JF7ngP5.png

 

[Nergal]

Soooooooo, according to OP; the secret of Doom is dat AMD made a smart move?
You little AMD fanboi you! (lol)

But really, isn´t this argument void?
Or AMD manipulated everyone and their evilness spreaded such they have an (seemingly) advantage.
Or AMD did some smart planning.

Results are the same. Which you could hardly call a secret.
(I am not going to mention NV here; isn´t that actually another topic all togheter?)

 

[RejZoR]

The fact AMD has been working on Mantle and GCN for ages, it's more of a long term smart planning than anything short term evil...

 

[R-T-B]

Just a hint, with OpenGL, NVIDIA also has it's own extensions and no one seems to have problems with that. But oh noes, Vulkan uses AMD specific goodies. Uh?

Does Doom use them?

 

[GoldenX]

I think they do, there is no reason they should not. Everybody has their own extensions, even Intel and SIS, (the SIS one is the most used of them all, s3tc).

 

[Dethroy]

The current situation is rather simple...

AMD = wider architecture
vs
Nvidia = narrower architecture with higher clocks

The thing is, it's hard to write software that scales as well with parallelism as it does with clocks.
That's basically a mirror of the same ol' same ol' GHz vs CPU cores battle we've all been watching for years.

 

[TheoneandonlyMrK]

The fact they have to make it work well on this wide gcn arch ,for consoles is likely to help the situation imho.
Shock horror tag line I'd expect more since Microsoft is definitely trying to push its devs to port to Uwp and dx12 from the Xbox side ,and it goes how it goes.

 

[arbiter]

Results are the same. Which you could hardly call a secret.
(I am not going to mention NV here; isn´t that actually another topic all togheter?)

well if you look at the image above, its like DX11/DX12. where as DX11 is lower end cpu and DX12 is higher end. mantle/vulkan was suposed to help lower end hardware get better performance but seems like that its back to square 1 and not doing that.

 

[FR@NK]

mantle/vulkan was suposed to help lower end hardware get better performance but seems like that its back to square 1 and not doing that.

Look at the graph again. On the systems with slower CPUs vulkan still gives you +10-15% over OpenGL.

 

[arbiter]

Look at the graph again. On the systems with slower CPUs vulkan still gives you +10-15% over OpenGL.

Yes it does give a boost but point i was making look at performance of amd x4 955, the radeon rx 480 loses a ton of performance, the 1060 loses very little compared to it.

Compare 6700k its 4% lose on vulkan for gtx1060 and ~40% for rx 480. When cpu comes more limiting factor the efficiency of nvidia drivers which has been seen they tend to be lower cpu usage in dx11 comes out.

 

[Melvis]

Bla Bla Bla as long as Im getting around 60FPS I dont care!

 

[ViperXTR]

something published recently from nvidia dev site, AMD had their own, nvidia had this one
https://developer.nvidia.com/reading-between-threads-shader-intrinsics
https://developer.nvidia.com/unlocking-gpu-intrinsics-hlsl

 

[ValenOne]

Hair works used DX11 Tessellation which has been apart of DX11 since day 1. as for PhysX, that is set to off in any game that was used it so it wasn't impacting anything in testing.

Vulkan USED to be mantle which was AMD closed proprietary API, AMD said they would open it up to all but as everyone expect that never happened and it stayed close source til AMD dropped the project and gave source to people that make OpenGL. However vulkan is still optimized and coded for AMD cards specifically so they get an advantage from it even though its not open source AMD has a good 2-3 or so year head start with it.


AMD's drivers are still miles behind, they have way more cpu overhead it seems when you are more cpu limited. Was a site that tested some older intel and amd cpu and under vulkan, RX 480 took pretty sizeable hit
Give size of image i will just provide a link for it, http://i.imgur.com/JF7ngP5.png

From https://developer.nvidia.com/dx12-dos-and-donts

On DX11 the driver does farm off asynchronous tasks to driver worker threads where possible - NVIDIA


NVIDIA DX11 drivers already using some of key DX12 speed up methods i.e. asynchronous tasks and multi-threads. This advantage was nullified under DX12.

something published recently from nvidia dev site, AMD had their own, nvidia had this one
https://developer.nvidia.com/reading-between-threads-shader-intrinsics
https://developer.nvidia.com/unlocking-gpu-intrinsics-hlsl

NVAPI has existed since the old "The Way It's Meant to be Played" program.

 

[RejZoR]

Vulkan is NOT Mantle. Assuming AMD has the edge because core idea behind Vulkan was Mantle is just absurd. Vulkan is to OpenGL what DX12 is to DX11. Mantle was just a core idea for both, but essentially they are entirely independent from Mantle itself.

 

[arbiter]

Vulkan is NOT Mantle. Assuming AMD has the edge because core idea behind Vulkan was Mantle is just absurd. Vulkan is to OpenGL what DX12 is to DX11. Mantle was just a core idea for both, but essentially they are entirely independent from Mantle itself.

http://www.pcworld.com/article/2894...ises-from-the-ashes-as-opengls-successor.html
http://www.legitreviews.com/amd-mantle-dead-known-vulcan-api-uses-mantle-technology-opengl_159339
Really want to say its not when it is. Pretty sure any reviewer will agree that it was mantle but when DX12 was announced it was DOA api hence it was killed and source given to kronos group which turned it in to vulkan.

 

[the54thvoid]

Vulkan is NOT Mantle. Assuming AMD has the edge because core idea behind Vulkan was Mantle is just absurd. Vulkan is to OpenGL what DX12 is to DX11. Mantle was just a core idea for both, but essentially they are entirely independent from Mantle itself.

http://www.pcworld.com/article/2894...ises-from-the-ashes-as-opengls-successor.html
http://www.legitreviews.com/amd-mantle-dead-known-vulcan-api-uses-mantle-technology-opengl_159339
Really want to say its not when it is. Pretty sure any reviewer will agree that it was mantle but when DX12 was announced it was DOA api hence it was killed and source given to kronos group which turned it in to vulkan.

And the judges result....

It looks like the first major announcement was made this morning when the Khronos Group announced that they will be using parts of the Mantle API to serve as the foundation for ‘Vulkan’
Read more at http://www.legitreviews.com/amd-man...-technology-opengl_159339#B16Lm9HsOvw9KUkl.99

AMD's Robert Hallock confirmed on a blog post that Mantle had, for the most part, been turned into the Khronos Group’s Vulkan API that would supersede OpenGL.

Quotes from articles. Kind of nails that one down to a win for @arbiter.

 

[RejZoR]

That's equally as stupid as saying S3 Metal was just a Glide. Core idea was the same, the actual thing wasn't. It's the same here. Mantle is NOT Vulkan.

Even Khronos own PR pamphlet says that:

The cross-vendor Khronos Group has chosen the best and brightest parts of Mantle to serve as the foundation for “Vulkan,” the exciting next version of the storied OpenGL API.

Foundation means core idea. How it operated aka close to metal. It doesn't mean this IS Mantle. Because it isn't. If it was, then it wouldn't work on NVIDIA. Obviously.

I mean, don't you find it peculiar how Vulkan and DX12 both appeared all of a sudden basically out of nowhere when AMD showcased practical use of Mantle in real games? It's because they took Mantle's core idea and added it to OpenGL and D3D11.

EDIT:
Corrected D3D12 to D3D11.

 

[medi01]

The new Nintendo thingy-ma-jig is Tegra.

Which isn't really a home console, but a handheld and will likely get next to no 3rd party support (again, lol). Nintendo knows how to play that though, in handhelds they reign supreme for, whahaha, nearly 30 years. (RIP Vita... )

(1080) beats all single GPU implementations by AMD GCN cards.

Well, that's hardly an argument given there is no new gen AMD card targeting that tier so far.
Given what 480 is vs 1060 (power consumption and chip size wise) I'm rather with you on this one, but still, let's wait and see the chips first.

Vulkan USED to be mantle which was AMD closed proprietary API,

Proprietary, but not closed. Unlike, cough...