DX12 GPU and CPU Performance Tested: Ashes of the Singularity Benchmark

I just want to download the benchmark myself and see what it looks like. I don't really care about tests that aren't my machine exactly.
 
Me too. But unfortunately they aren't giving it out to the public. Have to purchase the game first.
 
Looks like AMD has massive driver overhead in DX11... making it no wonder they just abandoned optimization(go figure with their bitching about gameworks) and pushed mantle.
 
Looks like AMD has massive driver overhead in DX11... making it no wonder they just abandoned optimization(go figure with their bitching about gameworks) and pushed mantle.

you know they abandoned optimization how? We should be complaining about gameworks, AMD doing it and then us turning on them is retarded.
 
This is all preliminary stuff. Both companies are scrambling to optimize performance, and this is exactly one test with one engine. Until there's a few real-world games out in the wild there's not a large enough sample size to draw any significant conclusions.
 
Looks like AMD has massive driver overhead in DX11... making it no wonder they just abandoned optimization(go figure with their bitching about gameworks) and pushed mantle.

Yep, sounds like AMD has decent HW, but their SW sucks. That's at least better than their APU side which has shitty HW too. In the end you have to create good HW and have optimized SW controlling it or you can't be competitive. AMD needs to pull their head out of their ass.

That all being said, hopefully AMD survives long enough to get another chance w/ dx12. I like competition. Under dx11 there is not any.
 
Very interesting. I hadn't realized until this video that there could be unique DX11 performance issues for AMD cards. I guess time will tell as a number of games come out that work on both DX11 and 12.
 
Me too. But unfortunately they aren't giving it out to the public. Have to purchase the game first.

I have purchased the game and I still can't run the benchmark. Looks like it's very selective.

On a different note, I'd like to know exactly why nVidia claims it's not representative of a DX12 game. The guy in the video just brushes it off without explaining it.


Edit: https://forums.geforce.com/default/...rs/nvidia-not-happy-ashes-of-the-singularity/

Their major sticking point seems to be its pre-alpha status and, as such, is not as optimised as the final game will be. Meaning that there will be better frame rates all around in the final release. I think that's fair enough too.
 
Last edited:
you know they abandoned optimization how? We should be complaining about gameworks, AMD doing it and then us turning on them is retarded.

Who is we? AMD fan club?

And your sentence makes no sense. Semantic made a valid point regarding the quality of DX11 drivers and AMD.

Also, what exactly did AMD do? What is this "it" you talk about. Is the "it" poorly optimized drivers? Or is "it" mantle? If "it" is mantle you are mistaken about AMD and DX12. There is this propaganda machine running as if they forced MS's hand, but mantle adaptation was so poor without AMD enticing devs to do so MS had no fire under their ass to make a low level API. This was all for the devs, and hopefully it will trickle down to the users...
 
I can't help but laugh at the use cases they've come up with. 980/390x paired with a core i3? All that is gonna benchmark is whose GPU can bang off of the ceiling set by the CPU the hardest.
 
I enjoyed watching this. Finally a presenter without a ugly ass Goatee
 
Man nothing about those results paints AMD in a particularly good light when it comes to the historical side of things. Their DX11 overhead is catastrophically bad and the Vishera CPU's are still loosing out to an i3. DX12 is something of a god send for them, those results are damn impressive.

All that noise that Nvidia made with their big DX11 overhead enhancements they introduced April last year clearly wasn't just marketing spin if this data is any sort of indication.
 
My take away is

Its everything I hoped Mantle was touted as trying to achieve. Ultimately for me being, I might get the gaming experience I wanted without having to build/rebuild/upgrade every few years, financially that's tough for most of us.

Kinda why I bought a fast dual core over a quad back in 07. It feels like very few games take advantage of extra cores let alone threads, it went to a nice GPU to get better frames. Only recently I can find a few like BF4 or CIV BE that might take advantage of mass core count. (CIV BE not so much... my take)

Maybe I'm not getting it but, looks like 2x the frame rate on current tech. That's gonna be (hopefully) an amazing time for us RTS guys, and when we see more of DX12 maybe the whole pc game market too :) !
 
Have no intentions of installing win10, wish they would sell a X12 patch for win7 users
 
I can't help but laugh at the use cases they've come up with. 980/390x paired with a core i3? All that is gonna benchmark is whose GPU can bang off of the ceiling set by the CPU the hardest.
Um the entire point of DX12/Vulkan is to extremely reduce CPU overhead, and they give raw numbers too, you're not going to suddenly pull twice the FPS and then make the statement you just did. It makes zero sense at all.


Man nothing about those results paints AMD in a particularly good light when it comes to the historical side of things. Their DX11 overhead is catastrophically bad and the Vishera CPU's are still loosing out to an i3. DX12 is something of a god send for them, those results are damn impressive.

All that noise that Nvidia made with their big DX11 overhead enhancements they introduced April last year clearly wasn't just marketing spin if this data is any sort of indication.

Worth noting Nvidia has modified their DX11 quite heavily. Forget what it's called but they have their own version of a Mantle esque thing that runs as a in place for Dx11. It was their response it to...it was whenever they had that "massive DX11 performance boost" driver way back when. If someone can remember the name of that feature that'd be great.
 
Worth noting Nvidia has modified their DX11 quite heavily. Forget what it's called but they have their own version of a Mantle esque thing that runs as a in place for Dx11. It was their response it to...it was whenever they had that "massive DX11 performance boost" driver way back when. If someone can remember the name of that feature that'd be great.

The multi-threaded rendering?
 
Worth noting Nvidia has modified their DX11 quite heavily. Forget what it's called but they have their own version of a Mantle esque thing that runs as a in place for Dx11. It was their response it to...it was whenever they had that "massive DX11 performance boost" driver way back when. If someone can remember the name of that feature that'd be great.

They've simply optimized their driver layer for DX11, said modifying is simply optimization (it's nothing particularly radical). The additional API is NVAPI, it's not quite the same as Mantle since it's not a full replacement but it's used for things like better SLI support and advanced rendering control at the development level, I know that DICE have support for it built into frostbite.

https://developer.nvidia.com/nvapi
 
Im not usually one for videos but this was very interesting.....AMD had up to 80% improvements on a couple of tests. I guess all those rumors of amd cards benefiting from dx12 more than nvidia were not exactly a pipe dream after all...so far so good:D
 
Man nothing about those results paints AMD in a particularly good light when it comes to the historical side of things. Their DX11 overhead is catastrophically bad and the Vishera CPU's are still loosing out to an i3. DX12 is something of a god send for them, those results are damn impressive.

All that noise that Nvidia made with their big DX11 overhead enhancements they introduced April last year clearly wasn't just marketing spin if this data is any sort of indication.

Look at reviews with actual DX11 games comparing the 980 and 390X. There is no nVidia driver advantage. The two cards trade blows. AMD's DX11 drivers are fine. They just obviously haven't optimized their DX11 drivers for this test. I would worry about it if you own an AMD card that's pre GCN once the game debuts. By then though I think anyone on a 5000/6000 series GPU will want to upgrade to play this game anyway.
 
Look at reviews with actual DX11 games comparing the 980 and 390X. There is no nVidia driver advantage. The two cards trade blows. AMD's DX11 drivers are fine. They just obviously haven't optimized their DX11 drivers for this test. I would worry about it if you own an AMD card that's pre GCN once the game debuts. By then though I think anyone on a 5000/6000 series GPU will want to upgrade to play this game anyway.

whoa whoa whoa! That opinion is WAY too reasonable and open.
 
Um the entire point of DX12/Vulkan is to extremely reduce CPU overhead, and they give raw numbers too, you're not going to suddenly pull twice the FPS and then make the statement you just did. It makes zero sense at all.




Worth noting Nvidia has modified their DX11 quite heavily. Forget what it's called but they have their own version of a Mantle esque thing that runs as a in place for Dx11. It was their response it to...it was whenever they had that "massive DX11 performance boost" driver way back when. If someone can remember the name of that feature that'd be great.

It's called marketing. The driver you are referring to didn't do any more than any other performance driver does. They basically had nothing to counter Mantle with so they made believe that driver was the same thing. DX11 is the bottleneck. No driver is going to fix the API. Don't confuse multi-thread rendering with multi-core rendering. The prior is what DX11 supports. The latter is what Mantle supported and so does DX12 (and Vulcan). Multi-thread rendering does offer performance improvements but they are quite modest.
 
Also, what exactly did AMD do? What is this "it" you talk about. Is the "it" poorly optimized drivers? Or is "it" mantle? If "it" is mantle you are mistaken about AMD and DX12. There is this propaganda machine running as if they forced MS's hand, but mantle adaptation was so poor without AMD enticing devs to do so MS had no fire under their ass to make a low level API. This was all for the devs, and hopefully it will trickle down to the users...

Relax, and learn2read. "It" is right there in the first part of his sentence, "complaining".
 
Look at reviews with actual DX11 games comparing the 980 and 390X. There is no nVidia driver advantage. The two cards trade blows. AMD's DX11 drivers are fine. They just obviously haven't optimized their DX11 drivers for this test. I would worry about it if you own an AMD card that's pre GCN once the game debuts. By then though I think anyone on a 5000/6000 series GPU will want to upgrade to play this game anyway.

I still find it a little concerning that without optimization AMD's DX11 numbers are so incredibly low on a piece of software that they've been collaborating on so heavily. Typically speaking if your underlying driver at it's core is well put together you shouldn't suffer a 50% performance deficit in pretty much anything provided something hasn't been purposefully broken somewhere.
 
Look at reviews with actual DX11 games comparing the 980 and 390X. There is no nVidia driver advantage. The two cards trade blows. AMD's DX11 drivers are fine. They just obviously haven't optimized their DX11 drivers for this test. I would worry about it if you own an AMD card that's pre GCN once the game debuts. By then though I think anyone on a 5000/6000 series GPU will want to upgrade to play this game anyway.

It's pretty funny that as we got more and more proof of how broken AMD drivers were in DX11 people are still in denial because they don't understand why some places in games are purely gpu limited while others are more affected by cpu.
 
Very interesting. I hadn't realized until this video that there could be unique DX11 performance issues for AMD cards. I guess time will tell as a number of games come out that work on both DX11 and 12.

reminds me NVs FX card line...

the issue here is we wont see many DX12 games for awhile and then Pascal will be out
 
Looks like if you want to play future DX12 games you should buy AMD but if you want to stick to playing older stuff just get nVidia.
 
Looks like if you want to play future DX12 games you should buy AMD but if you want to stick to playing older stuff just get nVidia.

ok that is funny

None of this matters until we have a released playable game that uses DX12 then both companies will hopefully do everything they can to beat each other.

Competition is good!
 
ok that is funny

None of this matters until we have a released playable game that uses DX12 then both companies will hopefully do everything they can to beat each other.

Competition is good!
Indeed. One game that is basically a tech demo at this point doesn't say anything about what actual performance will look like across the hardware. However, I think this situation is a good kick in the ass for NVIDIA. Hopefully this will cause them to step up their game again after getting too comfortable in their seat of market dominance, while at the same time fostering competition again that will be good for everybody.
 
I think it is funny how AMD owners have responded to this canned tech demo.
 
Indeed. One game that is basically a tech demo at this point doesn't say anything about what actual performance will look like across the hardware. However, I think this situation is a good kick in the ass for NVIDIA. Hopefully this will cause them to step up their game again after getting too comfortable in their seat of market dominance, while at the same time fostering competition again that will be good for everybody.

I think it should be pointed out that a party like Nvidia has already had access to much more actual DX12 software and certainly has internal benchmarks of their DX12 implementation. They would have much more internal awareness of how well a larger sample software size is behaving for them in terms of DX11->DX12.

Actually if we look back to another publicly available DX12 benchmark in Starswarm, which uses the same engine as this, that did not exhibit any performance regression issues - http://www.anandtech.com/show/8962/the-directx-12-performance-preview-amd-nvidia-star-swarm/5
 
So basically, if you're with AMD and want to play this game, you'll have to upgrade to Windows 10.
 
Back
Top