computerbaseAshes of the Singularity Beta1 DirectX 12 Benchmarks

[Dygaza]

With massive stuttering like that, AMD's benchmark results are invalidated and basically unplayable.

Stuttering? That is the effect of vsync

 

[PhonakV30]

With massive stuttering like that, AMD's benchmark results are invalidated and basically unplayable.

even if they fix it , still faster than GTX 980ti and still are validated! :sneaky:

 

[Erenhardt]

With massive stuttering like that, AMD's benchmark results are invalidated and basically unplayable.

It is not the case in the tests that run below max refresh rate of the display. Which makes it almost all scenarios that are worth mentioning from a GPU point of view.
High fps runs are good for CPU bottleneck testing anyway.

So, where it matter, in GPU testing, this is non-issue performance wise. It is interesting thou, how the fix will pan out.

 

[Bacon1]

Its going gold in a few weeks.. its very much a finished preview beta. And your last statement doesn't make any sense. Both nVidia and AMD have said that their most recent GPUs are fully DirectX12 compliant. They *are* DX12 GPUs.

Hell Nvidia even mentioned Async Compute as a feature of the 980 TI back in May 2015

http://www.geforce.com/whats-new/articles/introducing-the-geforce-gtx-980-ti

Still don't have working drivers for it 9 months later? Poor showing Nvidia

 

[Headfoot]

GTX980TI doesnt scale over GTX980. The limitation is not the compute performance.

Dude why do you keep saying this? Every graph, literally every single one, shows the 980 Ti getting better FPS than the 980 getting better FPS than the 970. It scales. It may not scale well, but it scales.

 

[Bacon1]

Update: hours before the release of this article we got word back from AMD. They have confirmed our findings. Radeon Software 16.1 / 16.2 does not support a DX12 feature called DirectFlip, which is mandatory and the solve to this specific situation. AMD intends to resolve this issue in a future driver update.

DirectFlip is a DX11.1 feature, and it used to turn off Aero when using a full screen application.

https://msdn.microsoft.com/en-us/library/windows/hardware/dn653329(v=vs.85).aspx

Funny enough anandtech mentioned it here:

http://www.anandtech.com/show/5630/indepth-with-the-windows-8-consumer-preview/14

I don't doubt there is something else they can work on, but I haven't seen any docs for DirectFlip in DX12.

I'm guessing they really meant swap chains in dx 12:

https://software.intel.com/en-us/articles/sample-application-for-direct3d-12-flip-model-swap-chains

 

[sontin]

Dude why do you keep saying this? Every graph, literally every single one, shows the 980 Ti getting better FPS than the 980 getting better FPS than the 970. It scales. It may not scale well, but it scales.

It is faster because the limitation happens not at the same performance level. It doesnt scale with the units it has over the other cards.

Here is another AMD sponsored game: http://pclab.pl/art68473-7.html
The GTX980TI is 1.47x faster than the GTX970 and 2,3x faster than the GTX960.

Here is the benchmark from Ashes with the Crazy setting: http://pclab.pl/art67995-13.html

The GTX980TI is only 1.36x and 2.03x faster.

And there are other benchmarks showing the same behaviour: http://www.guru3d.com/articles-pages/ashes-of-singularity-directx-12-benchmark-ii-review,7.html

Even in a huge CPU limit nVidia cards dont really improve with DX12. Obviously the DX12 path is broken for them.

 

[Headfoot]

It is faster because the limitation happens not at the same performance level. It doesnt scale with the units it has over the other cards.

Here is another AMD sponsored game: http://pclab.pl/art68473-7.html
The GTX980TI is 1.47x faster than the GTX970 and 2,3x faster than the GTX960.

Here is the benchmark from Ashes with the Crazy setting: http://pclab.pl/art67995-13.html

The GTX980TI is only 1.36x and 2.03x faster.

And there are other benchmarks showing the same behaviour: http://www.guru3d.com/articles-pages/ashes-of-singularity-directx-12-benchmark-ii-review,7.html

Even in a huge CPU limit nVidia cards dont really improve with DX12. Obviously the DX12 path is broken for them.

So it does scale. Just not perfectly, which has always been the case. Games that scaled perfectly have always been the unusual exceptions.

It would be interesting to look at each game and plot the scaling on a graph, and see where this falls. Picking one other random game isn't enough to draw the conclusion you're drawing. N=2 means no statistical significance.

The same goes for DX12 games altogether. We dont even have 1 game in full release yet though we are close. We'll need a solid handful before we can draw meaningful conclusions about where each card lines up on a generalized basis. I see people on the internet claiming "390x + async = 980 Ti" which is 100% too early to call and far too dependent on a single benchmark. And I doubt that will be the case anyways.

 

[Abwx]

Here is another AMD sponsored game: http://pclab.pl/art68473-7.html
The GTX980TI is 1.47x faster than the GTX970 and 2,3x faster than the GTX960.


The GTX980TI is only 1.36x and 2.03x faster.

If 1.47x is 100% scaling then 1.36x is 92.5% scaling, both numbers are comparable, beside the 970 is RAM speed and size gimped so why should it scale the same in all games..??.

 

[IllogicalGlory]

Even in a huge CPU limit nVidia cards dont really improve with DX12. Obviously the DX12 path is broken for them.

The 980 Ti does improve significantly on the Celeron and FX-4300 with DX12. Mind you, these settings are heavily GPU-bound regardless with a 980 Ti only putting in around 40 FPS on their best result. Not a great test IMO.

The DX12 results otherwise are extremely close to the DX11 ones. Can you provide a reason why NV should improve with DX12 in this scenario?

 

[Glo.]

It is faster because the limitation happens not at the same performance level. It doesnt scale with the units it has over the other cards.

Here is another AMD sponsored game: http://pclab.pl/art68473-7.html
The GTX980TI is 1.47x faster than the GTX970 and 2,3x faster than the GTX960.

Here is the benchmark from Ashes with the Crazy setting: http://pclab.pl/art67995-13.html

The GTX980TI is only 1.36x and 2.03x faster.

And there are other benchmarks showing the same behaviour: http://www.guru3d.com/articles-pages/ashes-of-singularity-directx-12-benchmark-ii-review,7.html

Even in a huge CPU limit nVidia cards dont really improve with DX12. Obviously the DX12 path is broken for them.

I suggest looking at how Fury X scales over R9 390X. Is DX12 path broken also for Fury X?

 

[Silverforce11]

There seems to be a Frametime issue with AMD cards : http://www.guru3d.com/articles-pages/ashes-of-singularity-directx-12-benchmark-ii-review,10.html

There's also some kind of mandatory DX12 feature not supported by AMD's actual drivers.

Pcper is looking into this at the moment.

AMD cards in Ashes DX12 can render more than 60 fps, but the frames output is tied to VSYNC, or whatever the rate of the monitor refresh.

It looks like they are not fully "Game Ready".

Tomshardware did a very indepth look at frame times:

http://www.tomshardware.de/ashes-of...rectx-12-dx12-gaming,testberichte-242049.html

Looks like their monitor is higher than 60hz.

 

[Elixer]

Those of you that want to do your own benchmarks, this game is on sale at steam ( http://store.steampowered.com/app/228880/ ) & Stardock's site ( https://store.stardock.com/product/ESD-SDE-W071 ) for $25.

 

[Bacon1]

Lol, even the fury x looses performance with async on at 1080!!!
I guess they don't support async in hardware either...

http://www.computerbase.de/2016-02/ashes-of-the-singularity-directx-12-amd-nvidia/3/

Fury X gains 11% with async on @ 1080p.

 

[Bacon1]

AMD cards in Ashes DX12 can render more than 60 fps, but the frames output is tied to VSYNC, or whatever the rate of the monitor refresh.

It looks like they are not fully "Game Ready".

Tomshardware did a very indepth look at frame times:

http://www.tomshardware.de/ashes-of...rectx-12-dx12-gaming,testberichte-242049.html

Looks like their monitor is higher than 60hz.

Ugh wish there was a way to filter out some of those, so hard to view the graph with everything on top of eachother.

Also the translation is amazing:

In Full-HD can also Radeon R9 380X convince still quite good, while the very high overclocked GeForce GTX 960 can not win a flowerpot.

 

[dacostafilipe]

Also the translation is amazing:

Actually, the translation is 100% correct ^^

 

[Phynaz]

Once you guys are done arguing and waving your e-peen around, do you want to discuss the real story of this release?

Functioning EMA is the thing here. You don't have to choose anymore, buy one of each brand. Mix different generations.

Everyone wins.

 

[Mahigan]

DirectFlip is a DX11.1 feature, and it used to turn off Aero when using a full screen application.

https://msdn.microsoft.com/en-us/library/windows/hardware/dn653329(v=vs.85).aspx

Funny enough anandtech mentioned it here:

http://www.anandtech.com/show/5630/indepth-with-the-windows-8-consumer-preview/14

I don't doubt there is something else they can work on, but I haven't seen any docs for DirectFlip in DX12.

I'm guessing they really meant swap chains in dx 12:

https://software.intel.com/en-us/articles/sample-application-for-direct3d-12-flip-model-swap-chains

FCAT is broken so the Guru3D results are wrong: http://www.extremetech.com/extreme/223654-instrument-error-amd-fcat-and-ashes-of-the-singularity

Basically AMD support WDM instead of DirectFlip. It reduces tearing and image anomolies. WDM is new to DX12. NVIDIA is still using the older technology.

 

[sontin]

The 980 Ti does improve significantly on the Celeron and FX-4300 with DX12. Mind you, these settings are heavily GPU-bound regardless with a 980 Ti only putting in around 40 FPS on their best result. Not a great test IMO.

Those settings are not "heavily GPU-Bound":

http://pclab.pl/art67995-15.html

With DX11 the GTX980TI is 10% faster, with DX12 only 27% faster than the GTX970. This card schould be ~47-50% better.

The DX12 results otherwise are extremely close to the DX11 ones. Can you provide a reason why NV should improve with DX12 in this scenario?

There doesnt exists any reason why nVidia cards dont see an improvement on a 2 core CPU. The DX11 optimization needs at least 4 threads:

http://www.anandtech.com/show/8962/the-directx-12-performance-preview-amd-nvidia-star-swarm/4

http://www.anandtech.com/show/9112/exploring-dx12-3dmark-api-overhead-feature-test/3

 

[Shivansps]

I dunno, looks totally worth it to me The game is in Beta and the feature will be used in Hitman, Fable Legends, Deus Ex: Mankind Divided, Battlefield 5, Gears of War Ultimate Edition and others.

Spoiler

The point whould be to compare Async ON/OFF... still im only seeing 10% on Fury X, thats nothing to call home about.

 

[Shivansps]

That only happened in their review. All other reviews don't show this FPS drop under 1080p with Async turned on.

That was what i was asking... do you know if all reviews run the same version?

 

[Mahigan]

The point whould be to compare Async ON/OFF... still im only seeing 10% on Fury X, thats nothing to call home about.

It's up to 20%, and that depends on how pervasive its usage is:

 

[Dygaza]

In response, why Fury X lost few fps with async in one test at 1080p:

Also remember that on higher framerates there is like +-2 fps variation to both directions between runs, simply because the benchmark is dynamic. So don't look the tests with that accuracy.

So sometimes you can have 2 fps better normal batches performance, and in the same run you lose a bit elsewhere.

 

[Bacon1]

In response, why Fury X lost few fps with async in one test at 1080p:

Also remember that on higher framerates there is like +-2 fps variation to both directions between runs, simply because the benchmark is dynamic. So don't look the tests with that accuracy.

So sometimes you can have 2 fps better normal batches performance, and in the same run you lose a bit elsewhere.

Yep, also unlike most benchmarks, this is actually playing the game and all the AI are dynamically fighting and doing different things, so each run is different.

 

[Abwx]

Those settings are not "heavily GPU-Bound":

http://pclab.pl/art67995-15.html

with DX12 only 27% faster than the GTX970. This card schould be ~47-50% better.

Taking account of the frequencies scaling under DX12 is roughly 95%, 33% more perfs for 40% more shaders (given the 5% lower frequency of the 980ti) while bandwith is not even scaled by 40%...

So what is your point..?.

That it should overscale..??