Ashes of The Singularity DX11 vs DX12 - R9 Nano

[mysticjbyrd]

https://www.youtube.com/watch?v=PYWdV-xnkdY

I saw this on reddit, and was really astonished by the insane increase in FPS, especially for the AMD cpu, so I decided to post this here.

TLDW:
R9 Nano
Intel CPU: 4690k @ 4.4GHz
AMD CPU: A10 7970k


INTEL
48% higher average FPS in DX12 over DX11

The maximum FPS in DX11 (47.5 fps) damn near became the minimum in DX12 (42.4 fps).


AMD
The AMD cpu was much worse in FPS in DX11, but was only 5 avg & minimum FPS lower in DX12.

The avg fps from DX11 (22 fps) nearly doubled in DX12 (42.9 fps).

The maximum in DX11 (35.1 FPS) was actually lower than the minimum in DX12 (38.8 FPS)

 

[ShintaiDK]

Its just the API overhead and better threading for the rendering part. Nothing new. Its covered multiple times before on sites. And remember AMD got terrible DX11 performance to begin with in AOTS.
http://www.computerbase.de/2015-10/...ashes-of-the-singularity-directx-11-1920-1080
http://www.computerbase.de/2015-10/...ashes-of-the-singularity-directx-12-1920-1080

Be careful not to confuse this with actual in game play that may be completely different due to game logic. Assuming the game will ever finish.

And dont use youtube as a source.

 

[mysticjbyrd]

Its just the API overhead and better threading for the rendering part. Nothing new. Its covered multiple times before on sites. And remember AMD got terrible DX11 performance to begin with in AOTS.
http://www.computerbase.de/2015-10/...ashes-of-the-singularity-directx-11-1920-1080
http://www.computerbase.de/2015-10/...ashes-of-the-singularity-directx-12-1920-1080

Be careful not to confuse this with actual in game play that may be completely different due to game logic. Assuming the game will ever finish.
.

I know that AMD lacks DX11 support, and focused on Mantle DX12. However, my point was more to illustrate the increase in performance from the weaker CPU. The intel setup got a ~50% increase in FPS, while the weaker AMD setup saw a ~100% increase. A lot of that increase from the AMD setup has to be from better utilization of the CPU, otherwise we would see similar scaling as the intel setup. Right?

I am not boasting AMD, but merely saying that DX12 seems to greatly increase in game performance for weaker CPUs. Though AMD might be the greatest benefactors here as they are generally the weaker of the two. I think we all know that Intel has the better CPUs by a long shot, but DX12 seems to make the bottleneck less prevalent for weaker CPUs, namely AMD.

I realize this is already known, but I thought the video was a good visual illustration, which is why I used youtube as the source.



Can you explain the second part? What is game logic?

 

[ShintaiDK]

DX12 is nothing but rendering essentially. If DX12 was the only thing used, you wouldn't have a game. You would just be watching a rendered video. Game logic=actual game including AI etc.

A lot of benches likes to use prescipted benchmarks. In this way you lower CPU usage because all you want to test is the GPU. A good example is Tomb Raider. Where a 800Mhz Haswell in the benchmark part is about equal to a 4900Mhz Haswell with 7950CF. In game however, there is a 400% difference.

Game logic is primary ST, so that part is not doing any favours for AMD.

And for AOTS:

Do you think DX12 helps AMD in any way here compared to Intel?

Using YouTube have shown a lot of bad things in the past with errors and faults at best, while completely misguiding at worst. And using a YouTube link from some hardcore AMD shill doesn't make it any better.

 

[frozentundra123456]

Amazing the extrapolations being made from one single so called demo that is really just an API test and not a game at all. And yes, I disregard any YouTube video as a source, especially one that starts and ends with a logo from the manufacturer.

 

[DrMrLordX]

Pooh pooh the results all you want. If we see enough real-world AMD users benefiting from DX12 like that, hey, good for AMD users. If we don't, well too bad guys.

Also, I would like to point out, that the graphic you (ShintaiDK) posted showed no minimum fps values and does not include a Kaveri APU.

 

[mysticjbyrd]

DX12 is nothing but rendering essentially. If DX12 was the only thing used, you wouldn't have a game. You would just be watching a rendered video. Game logic=actual game including AI etc.

A lot of benches likes to use prescipted benchmarks. In this way you lower CPU usage because all you want to test is the GPU. A good example is Tomb Raider. Where a 800Mhz Haswell in the benchmark part is about equal to a 4900Mhz Haswell with 7950CF. In game however, there is a 400% difference.

Game logic is primary ST, so that part is not doing any favours for AMD.

And for AOTS:

Do you think DX12 helps AMD in any way here compared to Intel?

Using YouTube have shown a lot of bad things in the past with errors and faults at best, while completely misguiding at worst. And using a YouTube link from some hardcore AMD shill doesn't make it any better.

No, but I am not saying it only helps weak AMD CPUs, but weak cpus in general.

Amazing the extrapolations being made from one single so called demo that is really just an API test and not a game at all. And yes, I disregard any YouTube video as a source, especially one that starts and ends with a logo from the manufacturer.

I don't think you know what an extrapolation is, as I merely summarized the results of the test.

Pooh pooh the results all you want. If we see enough real-world AMD users benefiting from DX12 like that, hey, good for AMD users. If we don't, well too bad guys.

Also, I would like to point out, that the graphic you (ShintaiDK) posted showed no minimum fps values and does not include a Kaveri APU.

WoW the AMD hate is strong here.

I am not a fanboy of either camp. I believe it's rather foolish to make decisions based on biased affiliations when we can rely on quantitative analysis.

 

[ShintaiDK]

No, but I am not saying it only helps weak AMD CPUs, but weak cpus in general.

2 options.

1. Either DX12 is used to lower power consumption.
2. Or DX12 is used to boost game logic or simply add even more on screen. (AAA games).

Option 2 means you end up with the same case as DX11 in terms of CPU requirement.

 

[mysticjbyrd]

2 options.

1. Either DX12 is used to lower power consumption.
2. Or DX12 is used to boost game logic or simply add even more on screen. (AAA games).

Option 2 means you end up with the same case as DX11 in terms of CPU requirement.

“In DirectX 12 Async Shaders supercharge work completion in a compatible AMD Radeon GPU by interleaving these tasks across multiple threads to shorten overall render time."

AMD then talked about the second brand new DirectX 12 feature i.e. Multi-Threaded Command Buffer Recording. Mentioning the immense improvements that this new feature will bring to gaming with DirectX 12. The 5 Key points about this feature are as follows:

1) Overhead is significantly reduced by moving driver and API code to any available CPU thread
2) The absolute time required to complete complex CPU tasks is notably reduced
3) Game workloads can be meaningfully distributed across >4 CPU cores
4) New “bandwidth” on the CPU allows for higher peak draw calls, enabling more detailed and immersive game worlds
5) All available CPU cores may now “talk” to the graphics card simultaneously


Once we move onto the Mantle and DirectX 12 results, AMD more than redeems itself. There are immense boosts to draw call throughput from start to finish on every processor tested, the largest boost coming from the FX 8350 where the R9 290X receives a frankly monumental boost to performance in the order of 1,600 per cent when single-core DX11 and DX12 scores are compared.
http://www.eurogamer.net/articles/digitalfoundry-2015-why-directx-12-is-a-gamechanger


According to Microsoft, one of the key problems in DX11 is thread utilization. Here’s a threading comparison between DX11 (top) and DX12 (bottom):

DX11 - DX12 CPU comparison

See how, in DX11, the entire workload is hanging on a single thread with extremely low utilization on the other threads? That’s a problem — with the kernel-mode driver running on the same thread as the game and the D3D layer, there’s just not much for the other threads to do. The second graph shows how, by splitting the workload more evenly, the game can hit much lower latencies. Better latencies translates directly into higher frame rates.

http://www.extremetech.com/gaming/1...ia-gpus-will-deliver-mantle-like-capabilities

 

[ShintaiDK]

And this is relevant how? Besides just copy/paste PR.

 

[frozentundra123456]

I am quite aware what an extrapolation is. All this hype about DX12 both for how it is magically going to transform AMD cpus and GPUs is not really extrapolation, more like wishful thinking.

 

[ShintaiDK]

I am quite aware what an extrapolation is. All this hype about DX12 both for how it is magically going to transform AMD cpus and GPUs is not really extrapolation, more like wishful thinking.

We got AOTS Alpha, Fable and a couple of others. And besides for AMD getting a much needed and awaited multithreaded driver there hasn't been any "magic". So yep, pretty much wishful thinking.

DX10 gave a 20% API reduction, that part disappeared faster than it could arrive as well.

 

[Flapdrol1337]

48% higher average FPS in DX12 over DX11

But how relevant is that increase? In ashes amd dx11 (compared to nvidia) performs much worse than usual.

Amd probably didn't bother optimising the ashes dx11 in their driver, since it'll have dx12.

 

[2is]

2 options.

1. Either DX12 is used to lower power consumption.
2. Or DX12 is used to boost game logic or simply add even more on screen. (AAA games).

Option 2 means you end up with the same case as DX11 in terms of CPU requirement.

Option 3: All of the above.

We are talking about PC games most of which have user adjustable settings you can tweak for the desired combination of performance/eye candy.

 

[TheELF]

Be careful not to confuse this with actual in game play that may be completely different due to game logic. Assuming the game will ever finish.

Single threaded game pretty much,getting choked up by graphics.

Dx12 gets rid of the really huge driver thread,and that's a good thing, but improving your framerate,not so much, unless you pair a tiny CPU with a monster GPU you won't get much out of dx12 in real life.

 

[Erenhardt]

So you guys say splitting huge driver thread into multiple threads will not give more benefit to amd strategy of MOAR COREZ?

I swear I remember some of you trying to show amd dx11 driver thread hogging OCed haswell core. Is that not the case now?

 

[frozentundra123456]

There may be some benefit, but lets wait for some benchmarks on real games, rather than basing the proposition on thinly veiled AMD PR in some you tube video.

 

[TheELF]

So you guys say splitting huge driver thread into multiple threads will not give more benefit to amd strategy of MOAR COREZ?

Look at the pics, the driver thread does not get split up, it disappears!
It (dx12) does not have more threads it has LESS!!
And it will give more benefit only if you could not get your GPU to 100% with dx11.

 

[ShintaiDK]

So you guys say splitting huge driver thread into multiple threads will not give more benefit to amd strategy of MOAR COREZ?

I swear I remember some of you trying to show amd dx11 driver thread hogging OCed haswell core. Is that not the case now?

Maybe you should just see what the results so far shows:

What do you see?

 

[Erenhardt]

Some numbers having no pattern what so ever that are aggregate of multiple workload bottlenecks.

We know multiple polished DX11 games that have their code exceptionally multithreaded. In those situations the main bottleneck is a single driver thread. CPU with weak single thread performance and many threads will be held back with most of its resources idle.

 

[ShintaiDK]

Slow cores stay slow, no matter if its DX12 or not.

 

[TheELF]

Some numbers having no pattern what so ever that are aggregate of multiple workload bottlenecks.

We know multiple polished DX11 games that have their code exceptionally multithreaded. In those situations the main bottleneck is a single driver thread. CPU with weak single thread performance and many threads will be held back with most of its resources idle.

No it's not,at least not only, there is also the main game thread that is always going to slow down a game on a cpu with slow single core performance, no matter how many cores you have or how many threads the game has,the main processing happens on one single thread.

 

[MajinCry]

Slow cores stay slow, no matter if its DX12 or not.

Yet more efficient code will run better, on both a slow core and a fast core, than less efficient code.

It's not a difficult concept.

 

[ShintaiDK]

Yet more efficient code will run better, on both a slow core and a fast core, than less efficient code.

It's not a difficult concept.

Neither is the benchmark graph with the CPUs listed.

 

[Fjodor2001]

So in short for DX11 vs DX12:

* DX12 favors multi core CPUs
* DX12 improves performance more for AMD CPUs than Intel CPUs
* DX12 will improve CPU efficiency in general, making it less likely that CPU performance will become the bottleneck. I.e. you can cope with a less powerful CPU, and better spend that money on a better GPU.