Unigine Engine Looks To Wasteland 2

[NomadDemon]

i tester unigine benchmarks on my ubuntu 11.04 64bit with radeon 4850 + blob

my max resolution is 1280x1024 and my fps was SAME on windows and on linux, about 50-60 [settings like high-ultra]

graphic was awesome, dont need teselation stuff

[Kano]

@ZedDB

http://www.phoronix.com/scan.php?pag...re_linux&num=1

One option that is not mention in the old article is:

aticonfig --cfl on

to enable a crossfire logo, but i never saw this. Usually i don't like logos, but to verify i would like to see that. I could only "verify" with 12-3 driver that with enabled cf it showed wrong colors, but no speed diff.

But did YOU really read your mentioned website correctly? There are only SLI results for OpenGL (which are pointless for Linux as SLI is not supported on Fermi there). There is NO OpenGL CF result just because there is no CF OpenGL profile. Everything is optimized for DX11 on AMD - absolutely nothing for OpenGL. So lets compare the speed differences of the single gpu cards @ 1080p (i would prefer fps, but hopefully the points do scale in a similar way, somebody from u should talk about that) - what you still miss is the Win -> Linux fps drop which are also a few fps for amd, for nvidia this is only a slight difference.

Card 	 DX11 	 OpenGL 	 (OpenGL/DX11)
HD 7970 	1467	1236	84,25%
HD 7950 	1300	1057	81,31%
GTX 580 	1105	1030	93,21%
HD 7870 	1059	906	85,55%
HD 6970 	973	688	70,71%
GTX 570 	943	881	93,43%
HD 7850 	940	790	84,04%
GTX 480 	918	886	96,51%
HD 6950 	877	610	69,56%
GTX 560 Ti 	811	742	91,49%
HD 6870 	785	564	71,85%
HD 7770 	774	627	81,01%
HD 5870 	773	577	74,64%
GTX 560 	764	645	84,42%
HD 6850 	669	477	71,30%
HD 7750 	508	429	84,45%
GTX 550 Ti 	489	421	86,09%

If you would have looked correctly then you should have seen that you lose about 14 to 30% depending on the AMD card used only by switching from DX11 to OpenGL (my HD5670 is more the 30% type). You lose a few % extra on ATI when you do you benchmarks on Linux. Lets look at Nvidia: you lose about 4 to 16% but you don't lose much when you switch from Win to Linux. That means switching to Linux costs you about 1/3 of the speed for AMD but only 1/6 with a SINGLE Nvidia card. As you can see that even if CF is officially supported on Linux by AMD you lose not only 1/3 of the speed, you loose MUCH more. If you are lucky your speed drops only by 2/3 from a DX11 CF solution to OpenGL Linux...

@binstream

The one and only way why Oilrush is seen as "fast" by some ppl is that absolutely no tessellation is used. Also you should release performance counts like number of objects/scene or whatever was reduced to improve speed. If the engine would be used the same way it would be slow as hell for most users without highend cards.

[Stebs]

@Kano
I don't really understand what you are "moaning" about. Currently the OpenGL drivers from Nvidia and especially AMD are not as optimized for Games as the DX11 drivers are, nothing new there.
And not really astonishing when you consider that most Games today use D3D (and OpenGL often used for "professional" applications) , but still it is a shame.
So where do you see the fault of the Unigine Engine in all this?
Until someone from Unigine tells us more, their DX11 path being more tweaked than their OpenGL path is pure speculation. It could as well be entirely a driver issue as mentioned before.

About tesselation: sure it is quite performance taxing, especially when used excessively like the Heaven _Benchmark_ does. So using it (and to what extend) will always be a question of cost/benefit. But this is true for every Engine.

[Kano]

First of all i have got nothing against Unigine, but some issues should be adressed. It is a bit similar to Rage issues with AMD drivers. If you want you can blame AMD for everything If you want to see crossplattform games using OpenGL 4/DX 11 features this should not result in such a performance penalty that even their own game does not use em. It is hard to believe that they can not reproduce the same results at their own office. Maybe Unigine could get some hints from AMD how to optimize the code in a way those cards work faster. Most likely AMD does that on the fly for DX11, maybe just rename the binary and try to bench again. This could be a bit tricky however because the "real" binary is executed by a launcher with the selected settings. I really want to see Linux games that use the latest OpenGL 4 features but i doubt that a bit even when a game will use this engine. A first step could be working together with AMD to get an OpenGL profile for CF - Nvidia has got SLI for that as well - sometimes it would be cheaper to add the same card again instead of buying a new one, but thats pretty useless currently for OpenGL - especially on Linux. All you can do is to get a fast card directly even when your board would support more cards.

[smitty3268]

First of all i have got nothing against Unigine, but some issues should be adressed. It is a bit similar to Rage issues with AMD drivers. If you want you can blame AMD for everything If you want to see crossplattform games using OpenGL 4/DX 11 features this should not result in such a performance penalty that even their own game does not use em. It is hard to believe that they can not reproduce the same results at their own office. Maybe Unigine could get some hints from AMD how to optimize the code in a way those cards work faster. Most likely AMD does that on the fly for DX11, maybe just rename the binary and try to bench again. This could be a bit tricky however because the "real" binary is executed by a launcher with the selected settings. I really want to see Linux games that use the latest OpenGL 4 features but i doubt that a bit even when a game will use this engine. A first step could be working together with AMD to get an OpenGL profile for CF - Nvidia has got SLI for that as well - sometimes it would be cheaper to add the same card again instead of buying a new one, but thats pretty useless currently for OpenGL - especially on Linux. All you can do is to get a fast card directly even when your board would support more cards.

I think it's pretty common for engines to favor nvidia or AMD over the other. That's why GPU hardware reviews generally include multiple tests (except for Phoronix, which uses 100 variations of Q3).

Anyway, out of curiosity, what happens if you disable Catalyst AI? I don't know if that is tweaked to work better on D3D11/Windows than Linux, or if it wouldn't make any difference.

[yogi_berra]

I'm a Wasteland 2 backer, and really hope they go with Unigine. I think it would be a nice fit for the type of game they're making. Wasteland 2 would also give Unigine a bit of publicity, and hopefully get more people interested in using it.

Imagine how bad Unigine will look if they don't use it.

They aren't making the choice of game engine on the engine itself, but on documentation of the engine, the tools that come with the engine, and how easy it is to integrate their tools with the engine. They're getting my money no matter what engine they use, but I would rather have them spend $100,000 for an engine that doesn't take a lot of effort to start scripting then to take a free license and have to spend a month rebuilding the tools just to get to the point of where they can start scripting.

[benmoran]

Well I agree with that. In the end it's up to the Inxile guys to determine if it's a good match for them. We'll just have to wait and see I guess. If they don't end up going with Unigine, it would interesting if they would explain exactly why they didn't. It's all just speculation now.

[binstream]

Binstream,

I just wanted to say thank you for supporting the open source drivers as well. With Oil Rush the performance is really impressive, all things considered. I finally had time to play past the second chapter, and it really is a lot of fun.

I'm a Wasteland 2 backer, and really hope they go with Unigine. I think it would be a nice fit for the type of game they're making. Wasteland 2 would also give Unigine a bit of publicity, and hopefully get more people interested in using it.

Thank you for your kind words!

[binstream]

And your tests show no extreme different performance with heaven between dx11 and opengl and on your systems you get better performance using crossfire with opengl?

There is no extreme difference in performance with DX11 and OpenGL

[Kano]

What attribute do you use for 30% less speed?