Quantum Break тест GPU (Gamegpu.com)

[Bacon1]

Game GPU Tests Then and Now

So the ones you tried to tell me where invalid and can't be compared?

https://forums.anandtech.com/threads/quantum-break-тест-gpu-gamegpu-com.2487804/#post-38494302


The ones that showed AMD being much faster in DX12 in the unpatched game compared to the current DX11 game?

Gotcha.

Fury X went from 62/68 -> 53/66
Fury Nano went from 55/60 -> 47/58
290x went from 47/52 -> 34 / 45 <-- That is worse AVG than previous MIN

 

[Sweepr]

So the ones you tried to tell me where invalid and can't be compared?

https://forums.anandtech.com/threads/quantum-break-тест-gpu-gamegpu-com.2487804/#post-38494302


The ones that showed AMD being much faster in DX12 in the unpatched game compared to the current DX11 game?

Gotcha.

Yes, the ones where old vs new (steam version) comparisons might not be valid - but you're pretending to be confused about the fact that we can still compare how the same cards fare relative to each other then and now (same versions) - that ones.

 

[Bacon1]

Game GPU Tests Then and Now:

Yes, the ones where old vs new (steam version) comparisons are invalid -

So you are using the invalid comparisons when it suits you but otherwise it's not ok to? Interesting.

 

[Bacon1]

Doesn't justify a massive performance gain in DX11 relative to DX12 - moving up relative to competing cards. Also you don't need to post about Async Compute drivers and Maxwell doing bad in DX12 in every single thread, it's getting repetitive.

You brought it up before I did. I never mentioned Async compute.

Yes, I thought DX12 was supposed to improve or (worst case scenario) not cripple the experience for any vendor, but that doesn't seem to be the case here.

Actually it does very well in Forza Horizon 3 and other non Gaming Evolved DX12 titles (few exceptions) - but I guess those are not 'real DX12 titles' in your view.

 

[Sweepr]

So you are using the invalid comparisons when it suits you but otherwise it's not ok to? Interesting.

Comparing the FPS of the cards is indeed wrong because it's not the exact same scene tested and they might have used different settings, but we can still compare how NVIDIA x AMD fare against each other then and now.

Hint, the Maxwell cards that you criticize so much seem to be doing fine, what do you think?

GTX 780 equal to R9 280X before, now faster than R9 290X.
GTX 970 equal to R9 380X before, now beating RX 470.
GTX 980 Ti slightly behind Fury X before, now faster.

 

[Bacon1]

Comparing the FPS of the cards is indeed wrong because it's not the exact same scene tested and they might have used different settings, but we can still compare how NVIDIA x AMD fare against each other then and now.

Not when its obvious that one vendor is losing massive amounts of performance. They are either invalid, or valid. You can't cherry pick only the parts of the results you like.

 

[Bacon1]

Hint, the Maxwell cards that you criticize so much seem to be doing fine, what do you think?

Sounds like they don't handle DX12 from what we've seen. Yet another game they gain by going back to DX11.

AMD can do DX12 and gains from it, thats why your ordering is different now. The 290x lost a ton of FPS.

Fury X went from 62/68 -> 53/66
Fury Nano went from 55/60 -> 47/58
290x went from 47/52 -> 34 / 45 <-- That is worse AVG than previous MIN

So the unpatched DX12 version was faster than the latest DX11 version. Once we have newer side/side testing of latest DX12 vs DX11 we'll have better info. Hopefully they keep their word and make both Win Store and Steam versions both DX11/DX12.

 

[Sweepr]

AMD can do DX12 and gains from it, thats why your ordering is different now. The 290x lost a ton of FPS.

The fact that AMD cards gain more doesn't justify a massive performance deficit for Maxwell in DX12 relative to DX11. But I'm sure the answers will be here in the coming days.

 

[SPBHM]

my understanding is that they were forced to launch as DX12 by MS, they feel more competent and think it's a better choice with DX11, that's why the non MS app store release is DX11 (and works on windows 8 and 7),
this version should fix lots of bugs and performance, it will be interesting to see a proper comparison if in any scenarios the DX12 version is faster.

 

[Carfax83]

What other DX12 titles does Maxwell gain over DX11?

The greatest of them all, AotS.. And whether Maxwell gains in DX12 over DX11 has nothing to do with Maxwell's performance in DX12, and everything to do with the particular implementation of DX12.

Developers have to be skillful enough to overcome NVidia's highly optimized DX11 driver, which mimics the low level optimizations that DX12 affords to a certain degree. With AMD, this isn't an issue because AMD's DX11 driver has terrible CPU overhead, so of course they always gain with DX12 since they are starting at a much lower baseline than NVidia.

 

[Carfax83]

my understanding is that they were forced to launch as DX12 by MS, they feel more competent and think it's a better choice with DX11, that's why the non MS app store release is DX11 (and works on windows 8 and 7),
this version should fix lots of bugs and performance, it will be interesting to see a proper comparison if in any scenarios the DX12 version is faster.

This pretty much. Remedy, being a small studio shouldn't really mess with DX12. DX12 is supposed to be for advanced developers with great technical skill and ability. This just goes to show how much DX12's performance depends on the quality of the implementation...

 

[Bacon1]

The greatest of them all, AotS.. And whether Maxwell gains in DX12 over DX11 has nothing to do with Maxwell's performance in DX12, and everything to do with the particular implementation of DX12.

That wouldn't count though, as its a GE game according to Sweeper. He was specifically talking about non GE games

Actually it does very well in Forza Horizon 3 and other non Gaming Evolved DX12 titles (few exceptions) - but I guess those are not 'real DX12 titles' in your view.

Developers have to be skillful enough to overcome NVidia's highly optimized DX11 driver, which mimics the low level optimizations that DX12 affords to a certain degree. With AMD, this isn't an issue because AMD's DX11 driver has terrible CPU overhead, so of course they always gain with DX12 since they are starting at a much lower baseline than NVidia.

If Nvidia's drivers are so great and AMD's so bad, why are many DX12 gains similar between the two then?

I mean in your example:

1080 went from 56 -> 70 fps in DX12.
Fury X went from 40 -> 50
980 TI went from 37 -> 53, even MORE than the Fury X which apparently has worse DX11 drivers...

Thats impossible according to you, as Nvidia's DX11 drivers are so much better than AMDs.

 

[Sweepr]

That wouldn't count though, as its a GE game according to Sweeper. He was specifically talking about non GE games

Stop putting words in my mouth, it's getting really embarassing for you.

Actually it does very well in Forza Horizon 3 and other non Gaming Evolved DX12 titles (few exceptions) - but I guess those are not 'real DX12 titles' in your view.

Used FH3 as an example but didn't say all GE titles see poor results from Maxwell @ DX12. And AotS improved a lot from launch till now on the NVIDIA side.

 

[Bacon1]

Stop putting words in my mouth, it's getting really embarassing for you.

I'm not? You've called AOTS GE title. I even quoted you in my original post, so no idea why you quoted it again as well.

Thank you very much for proving my point. Ignoring the obvious GE outliers AotS (which ran at DX12 on AMD cards, unlike Deus EX here) and Hitman (most biased title ever) - GTX 1080 is easily above >20% faster in neutral titles. So yes, the slower and more power hungry Fury X coming close is the exception and not the rule.

Btw please stop with the personal attacks. You've had quite a few in this thread against me.

 

[Sweepr]

I'm not? You've called AOTS GE title. I even quoted you in my original post, so no idea why you quoted it again as well.

Why exactly are you quoting a post from another (months old) thread? I don't see how your thread derail (and stalking) is relevant to this discussion.

What's wrong about calling AotS an AMD title?

http://www.amd.com/en-us/markets/game/featured/ashes-of-the-singularity
http://www.guru3d.com/news-story/amd-announces-directx-12-game-engine-developer-partnerships.html

 

[Bacon1]

Why exactly are you quoting a post from another (months old) thread? I don't see how your thread derail is relevant to this discussion.

You said I had misquoted you, I was just showing that you've called it a GE title so how is it a misquote?

The one derailing the thread is you. Look at your post it has much more space than mine and is completely off topic. You were the one bringing up other titles into this thread.

 

[Sweepr]

You said I had misquoted you, I was just showing that you've called it a GE title so how is it a misquote?

Not sure if you're confused or not but I refer to this part:

He was specifically talking about non GE games

Which is wrong of course, because I didn't say (all) GE titles should perform worse on Maxwell @ DX12 - that's another ill-intentioned post putting words in my mouth.

I'll let you continue your monologue of derailing, callouts and misquotes now. Back to topic.

 

[Bacon1]

Not sure if you're confused or not but I refer to this part:



Which is wrong of course, because I didn't say (all) GE titles should perform worse on Maxwell @ DX12 - that's another ill-intentioned post putting words in my mouth.

Well then maybe just say exactly what you want next time instead of being completely ambiguous. You've shown nothing but contempt for AOTS in the past so why would I expect that to be the one game you now think is valid to use?

Can you please stay on topic now and stop making personal attacks against me?

 

[Carfax83]

If Nvidia's drivers are so great and AMD's so bad, why are many DX12 gains similar between the two then?

I mean in your example:

1080 went from 56 -> 70 fps in DX12.
Fury X went from 40 -> 50
980 TI went from 37 -> 53, even MORE than the Fury X which apparently has worse DX11 drivers...

Thats impossible according to you, as Nvidia's DX11 drivers are so much better than AMDs.

This isn't really an apples to apples comparison like you'd like to portray. AotS is a Radeon optimized title.. The Nitrous engine is specifically optimized for GCN hardware (similar to how UE4 is highly optimized for NVidia) to a much higher degree than other engines. So basically, AMD already gets a leg up over other IHVs courtesy of Oxide's efforts, and this includes DX11.

As for why Maxwell sees a larger boost than Pascal, one can presume that it's because Pascal is just a much more efficient architecture than Maxwell, so DX12 doesn't really add as much in terms of overall efficiency and throughput compared to DX11..

 

[Azix]

I'm wondering if they toned down some of the graphics (did them some more efficient way) for the steam release.

This pretty much. Remedy, being a small studio shouldn't really mess with DX12. DX12 is supposed to be for advanced developers with great technical skill and ability. This just goes to show how much DX12's performance depends on the quality of the implementation...

They seem to have implemented some rather complex tech in quantum break besides dx12. I guess its not the deities of game programming that make good use of everything. strangely enough, they also seem to have built an engine just or that game. From what you guys have said about engine development and dx12 game development, one would think they could do neither. Building dx12 alongside dx11 with such low level programmers in a small studio. Oh dear. magic.

The Northlight Engine[1][2] is the in-house game engine created by Remedy Entertainment. The first game produced with the Northlight Engine was Quantum Break, an Xbox One and Windows 10 exclusive title.

History
Remedy Entertainment's previous endeavors, Max Payne and Alan Wake were both designed with in-house engines created by Remedy. The engine used in Alan Wake, dubbed "The Alan Wake Engine", was created using the Havok physics component.[3]

For Quantum Break, Remedy created a completely new engine, dubbed the "Northlight engine". According to creative director Sam Lake, the Northlight Engine was an evolution of their storytelling process. The Northlight Engine can purportedly capture facial motion capture using surface capture and 4D scanning better than previous engines developed.[1] Remedy’s Graphics Programmer, Ville Timonen, states that Quantum Break is the second game to utilize DirectX 12, which was implemented in the Northlight Engine.[4]

Early in the process of the Engine's use, the development team encountered problems with refining A.I. and fidelity issues[5], but argued early demo footage used to demonstrated was not fully implemented at the time.

http://quantumbreak.wikia.com/wiki/Northlight_Engine

they need to stop defying the laws of game development defined by you guys. All these engines they keep building.

This isn't really an apples to apples comparison like you'd like to portray. AotS is a Radeon optimized title.. The Nitrous engine is specifically optimized for GCN hardware (similar to how UE4 is highly optimized for NVidia) to a much higher degree than other engines. So basically, AMD already gets a leg up over other IHVs courtesy of Oxide's efforts, and this includes DX11.

As for why Maxwell sees a larger boost than Pascal, one can presume that it's because Pascal is just a much more efficient architecture than Maxwell, so DX12 doesn't really add as much in terms of overall efficiency and throughput compared to DX11..

two can play that game. AOTS is an nvidia biased game as more time was spent making it run better on nvidia cards than AMD cards. The nitrous engine was specifically optimized for nvidia GPUs since nvidia was given source code access to the engine and allowed to offer inputs that benefited their performance. There.

 

[Carfax83]

From what you guys have said about engine development and dx12 game development, one would think they could do neither. Building dx12 alongside dx11 with such low level programmers in a small studio. Oh dear. magic.

I don't think you've been reading what I've been saying. I never said Remedy couldn't do DX12 at any time. What I've been saying, is that DX12 should only be used by advanced developers that have the skills to implement it properly. Judging by the benchmarks, it's evident that Remedy are much better with DX11, than they are with DX12..

these guys need to stop defying the laws of game development defined by you guys. All these engines they keep building.

I don't know if you're deliberately misinterpreting what I've been saying, but I haven't said anything about Remedy not being able to make their own engine, or even implement DX12.

two can play that game. AOTS is an nvidia biased game as more time was spent making it run better on nvidia cards than AMD cards. The nitrous engine was specifically optimized for nvidia GPUs since nvidia was given source code access to the engine and allowed to offer inputs that benefited their performance. There.

NVidia required more optimization because the game engine was designed from a fundamental level to exploit GCN hardware. This includes Mantle optimization as well.. Nitrous engine was first demoed using Mantle, before it even had DX12. Also, both vendors had access to the source code..

And not to mention, the early benchmarks for DX12 on NVidia hardware were terrible for AotS. That's what kicked off the whole "NVidia can't do DX12" sentiment in fact.. Only after numerous patches and driver updates did the performance get to where it is now. And I will admit, that NVidia had some work to do with their DX12 driver as well, so it wasn't just Oxide's fault as to why DX12 performance was so low on NVidia hardware.

 

[nurturedhate]

This thread may very well contain some of the most overwhelmingly biased posts this forum has ever seem. Please, try to hide it better people. This is sad.

 

[tviceman]

The huge gains Nvidia is now getting under DX12 in AoTS completely disproves and puts to rest the theory that Pascal can't do DX12. It looks like games built with good all around DX12 support can net similar gains on either AMD or Nvidia.

 

[Unreal123]

The huge gains Nvidia is now getting under DX12 in AoTS completely disproves and puts to rest the theory that Pascal can't do DX12. It looks like games built with good all around DX12 support can net similar gains on either AMD or Nvidia.

Even if you compare Fury X performance on DX12 it still cannot beat GTX 980 Ti on DX11 in this game.

 

[frozentundra123456]

This thread may very well contain some of the most overwhelmingly biased posts this forum has ever seem. Please, try to hide it better people. This is sad.

Seems about par for the course in this forum.