PCPERDX12 GPU and CPU Performance Tested: Ashes of the Singularity Benchmark

[LTC8K6]

So the 980ti overclock capability will keep it well in front then?

 

[DarkKnightDude]

I'd like to see how a Fury X changes in performance in this game going from DX11 to DX12.

 

[Silverforce11]

I'd like to see how a Fury X changes in performance in this game going from DX11 to DX12.

All the performance charts were posted above, just scan through the thread. You will find 1080, 1440 and 4K.

 

[cmdrdredd]

So the 980ti overclock capability will keep it well in front then?

Well, the results from the German site show different results than the others I looked at. 1080p and 1440p performance dropped between DX11 and DX12 on pcper, but on computerbase.de the performance went up. However from what I see, the computerbase.de results are without MSAA. MSAA is what seems to kill Nvidia performance in this particular DX12 benchmark.

 

[cmdrdredd]

I'd like to see how a Fury X changes in performance in this game going from DX11 to DX12.

This post I made has those numbers.

http://forums.anandtech.com/showpost.php?p=37636776&postcount=6

 

[shady28]

I'd like to see how a Fury X changes in performance in this game going from DX11 to DX12.

Well in DX 11 a GTX 960 beats fury X, while in DX 12 Fury X beats the 980 Ti. And of course, Oxide said all of their code was reviewed etc etc.

Also there is a huge inconsistency in how the 980 Ti gained performance, and all other Nvidia cards lost performance going from DX 11->DX 12.

Unless we're ready to accept that as a new reality, I think this game has serious issues with optimization on multiple levels. I have never seen a shipping DX11 game where a Fury or 390X got clobbered by a 960.

 

[Silverforce11]

So the 980ti overclock capability will keep it well in front then?

Depends on what boost clocks the 980Ti sample here reached:

If its reference, reviews including AT shows it reaches 1.2ghz boost.

Assuming an OC to 1.5ghz boost or 25% OC and 100% scaling, it will go from 36.8 to 46fps.

Fury X has about 10% OC without +vcore. It goes from 41.4 to 45.5fps. Thats for DX12.

In DX11, the 980Ti will jump to 49fps.

Here's what nobody has mentioned, does the game look different at all in DX11 vs DX12 mode? What about the thousands of dynamic light sources (each gunfire is a dynamic light, each explosion, trails, etc) which the devs said wouldn't be possible without DX12?

 

[cmdrdredd]

Well in DX 11 a GTX 960 beats fury X, while in DX 12 Fury X beats the 980 Ti. And of course, Oxide said all of their code was reviewed etc etc.

Also there is a huge inconsistency in how the 980 Ti gained performance, and all other Nvidia cards lost performance going from DX 11->DX 12.

Unless we're ready to accept that as a new reality, I think this game has serious issues with optimization on multiple levels. I have never seen a shipping DX11 game where a Fury or 390X got clobbered by a 960.

That or it reflects the current state of Windows 10 drivers for this engine.

One of two things happened if I had to guess. Either Nvidia didn't put any or very little optimizations in the driver for DX12 or this game and AMD did, or we are looking at raw performance numbers from both without optimizations at all. Just because Oxide says they provided the code for Nvidia and AMD to go through and optimize for doesn't mean they did.

 

[Silverforce11]

I have never seen a shipping DX11 game where a Fury or 390X got clobbered by a 960.

Project Cars with Rain. The 960 clobbers all AMD GPUs, also destroys Kepler Titan as well.

 

[Silverforce11]

Crosspost, lets keep DX12 talk in this thread:

It is unusual but I think right now we are seeing raw GPU performance with little to no optimizations in the driver or maybe AMD has the optimizations and Nvidia doesn't. Nvidia is clearly behind in at this point when it comes to this particular game.

Oxide clearly said NV did optimized the game. They even listed an example.

For example, when Nvidia noticed that a specific shader was taking a particularly long time on their hardware, they offered an optimized shader that made things faster which we integrated into our code.

When they had source code for over a year, lack of optimization is no excuse.

 

[cmdrdredd]

Crosspost, lets keep DX12 talk in this thread:



Oxide clearly said NV did optimized the game. They even listed an example.



When they had source code for over a year, lack of optimization is no excuse.

When there is no game yet maybe Nvidia didn't put much work into it? It is pre-alpha right? They might have just been lazy.

Oxide doesn't know what Nvidia did with their driver either, they offered code for the game but that doesn't mean they did any work in their drivers for it yet. Nvidia usually releases a game ready driver for new games right at release. I don't know how long they hold those optimizations out.

 

[Silverforce11]

They did have time to release an official reviewer's guide to the media.. they should have had ample opportunity to optimize it. Instead of attacking the game for not being representative of DX12 gaming... which prompted a correction from Oxide.

This reminds me of Tessellation. Recall how NV pretty much said Tessellation wasn't important. Until Fermi. Then it's all about Tessellation.

I'm expecting the same with Pascal. It's all about DX12 & VR post-Pascal.

 

[tential]

Overreacting doesn't begin to describe the conclusions being drawn from this benchmark.

 

[Azix]

When there is no game yet maybe Nvidia didn't put much work into it? It is pre-alpha right? They might have just been lazy.

Oxide doesn't know what Nvidia did with their driver either, they offered code for the game but that doesn't mean they did any work in their drivers for it yet. Nvidia usually releases a game ready driver for new games right at release. I don't know how long they hold those optimizations out.

http://www.nvidia.com/download/driverResults.aspx/89124/en-us

Game Ready
Best gaming experience for Ashes of the Singularity preview

It is pretty strange that performance drops. Nvidia must know why since they have the code and tools. They just aren't being straight

 

[bystander36]

From what I read and I cannot verify any of this, is that Nvidia has some driver voodoo going on for DX11 in general to help with CPU bottlenecks and relieve some of that overhead. AMD doesn't have the same thing going on in their drivers which is why DX11 is such a wide margin in favor of Nvidia. When we go to DX12 there's very little driver optimization right now, if any at all. So it relies on the raw performance of the GPU at the moment. We are simply seeing that AMD's GPUs are hindered by a certain amount of driver overhead in DX11 without game specific optimizations to help out any.

While I believe that may be true to some extent, but this is much more massive than you'd see in other DX11 games which doesn't support Mantle. I'd have to believe AMD just decided not to try and do anything to help them develop for DX11 on this game.

 

[cmdrdredd]

http://www.nvidia.com/download/driverResults.aspx/89124/en-us



It is pretty strange that performance drops. Nvidia must know why since they have the code and tools. They just aren't being straight

Has anyone tried the older driver for comparison? This is why I wish we could get this benchmark. I want to try stuff like that and see what happens.

They did have time to release an official reviewer's guide to the media.. they should have had ample opportunity to optimize it. Instead of attacking the game for not being representative of DX12 gaming... which prompted a correction from Oxide.

This reminds me of Tessellation. Recall how NV pretty much said Tessellation wasn't important. Until Fermi. Then it's all about Tessellation.

I'm expecting the same with Pascal. It's all about DX12 & VR post-Pascal.

Their reviewer guide is a joke. It consisted of telling people to ignore MSAA results. At 1080p you should be able to run MSAA on a 970/980/980ti

 

[cmdrdredd]

double post.

 

[3DVagabond]

From what I read and I cannot verify any of this, is that Nvidia has some driver voodoo going on for DX11 in general to help with CPU bottlenecks and relieve some of that overhead. AMD doesn't have the same thing going on in their drivers which is why DX11 is such a wide margin in favor of Nvidia. When we go to DX12 there's very little driver optimization right now, if any at all. So it relies on the raw performance of the GPU at the moment. We are simply seeing that AMD's GPUs are hindered by a certain amount of driver overhead in DX11 without game specific optimizations to help out any.

I've read that it's easier with nVidia drivers to do optimizing. With AMD each game requires very specific optimizations but with nVidia the optimization is much more generic and can be implemented in a more general way.

 

[cmdrdredd]

While I believe that may be true to some extent, but this is much more massive than you'd see in other DX11 games which doesn't support Mantle. I'd have to believe AMD just decided not to try and do anything to help them develop for DX11 on this game.

It could be that AMD ignored DX11 for this since they knew it would support DX12 and they wanted to shine in that. I bet as time goes on things will smooth out in DX11 for AMD and they will have good performance there too with this title.

 

[guskline]

I'm anxious to test the "Ashes of the Singularity" benchmark on my rigs below GTX980TI SC vs R9290s in CF

 

[shady28]

Well here's a GTX 770 with an i5-3570K eating a Titan X / i7-5960X's for lunch.

I find it hard to believe any of these benchmarks are meaningful.

 

[cmdrdredd]

Well here's a GTX 770 with an i5-3570K eating a Titan X / i7-5960X's for lunch.

I find it hard to believe any of these benchmarks are meaningful.

After looking closely did you notice the Titan X wasn't running fullscreen? I wonder if that has an effect on the performance results. Also the Titan X was a full system test and the 770 a CPU test. Settings are not identical. Dunno what to say

 

[shady28]

Well here's a GTX 770 with an i5-3570K eating a Titan X / i7-5960X's for lunch.

I find it hard to believe any of these benchmarks are meaningful.


After looking closely did you notice the Titan X wasn't running fullscreen? I wonder if that has an effect on the performance results. Also the Titan X was a full system test and the 770 a CPU test. Settings are not identical. Dunno what to say

Should be fixed now.

Source is here : http://wccftech.com/ashes-singularity-alpha-dx12-benchmark/

 

[shady28]

After looking closely did you notice the Titan X wasn't running fullscreen? I wonder if that has an effect on the performance results. Also the Titan X was a full system test and the 770 a CPU test. Settings are not identical. Dunno what to say

The resolution on the Titan is also borked. This is what they said :

"It should be noted that for the first time here we encountered a bug preventing us from running the benchmark at fullscreen, so we had to opt for windowed at 1902×1033."

 

[shady28]

Here's another one. 780Ti / i7-3770 gets obliterated by 770 / i5-3570K.

The only explanation I can think of for this kind of stuff is that Nvidia has settings and optimizations in their DX12 drivers that are on for some cards, but not others.