AMD Catalyst Linux OpenGL Driver Now Faster Than Windows Driver In Some Tests

[Kano]

My own Unigine Heaven tests had been similar results a few years ago. But the main difference was CF or SLI support. AMD has no CF Heaven profile for Windows and with Linux it had rainbow colors first with active CF. I think CF D3D scaled well, better than SLI with my GTX 295, but there it was a DX 10/OpenGL 3 test only. Btw. without PTS look at the settings as the defaults are different between the OSs. AMD really only fixes OpenGL if they are under pressure, that lead to a bad start for Rage. The needed drivers had been missing on launch and the beta drivers had problems with some cards. I reported rendering problems with pointsprites/openscenegraph and it took about 11 months for a fix. It was just ignored as they thought only an unimportant benchmark app used it, but then somebody found the issue with Flightgear as well, based on OSG too. It is a bit weird that AMD considers source engine speed unimportant, at least all gamers working for AMD seem to use Windows only, otherwise somebody from inside would debug the issues with more vigor.

[curaga]

So are you suggesting D3D is harvesting energy (performance the hardware is incapable of) from out of nothing and beating the law of conservation of energy?

No. Read my post again. I said D3D does not reach "full performance" either, so it's useless if you're looking for "what my hw can do".

[yakman2020]

I'd be interested in seeing some real-world performance comparisions between Direct3D 11 Windows games vs OpenGL Linux games (via proprietary drivers).

Actually, I don't think you would see a lot of difference. Under the covers D3D and OGL in windows use a common pipeline. So long as the paths with that pipeline were similar, the pref would be pretty much the same.

[iznogood]

No, it doesn't. D3D is sufficiently different from both GL and HW that it doesn't tell that, not even talking about differently built engines for each.

You want to know what the hw can do? Look at specs and low-level microbenchmarks.

I don't understand this argument.
I buy a piece of hardware and I want the best possible performance. I don't care if it's on DX or opengl. So we compare opengl with the fastest implementation on windows. If opengl does not perform well under windows why should I care especially since it is not used anywhere? It will only hide the real problem which is the fact that mesa is not well optimized for several reasons and still can improve.

[Styromaniac]

Probably the best comparison would be with Phoronix tests pitting AMD and Nvidia against each other on Linux. For curiosity sake, benchmark Serious Sam 3 or other multi API supporting Windows game that's also available on Linux and switch between renderers on Windows. So Catalyst OpenGL is faster with Linux than Windows, but is Nvidia still faster than AMD concerning Linux? What about Gallium3D Nine vs Windows benchmarks? And no, I'm not confusing Gallium3D Nine as having anything to do with OpenGL, but if anything AMD helps Linux destroy Microsoft in any use case for gamers, that's worth remembering here.

Though, correct me if I'm wrong, but I think we already have all the information we need from the Linux side thanks to Michael and we can take that info to make buying decisions based on our needs. Or... I'm wrong because certain benchmark tests need to be revisited with newer drivers.

[artivision]

Phoronix: AMD Catalyst Linux OpenGL Driver Now Faster Than Windows Driver In Some Tests

Earlier this week I showed benchmarks of AMD's incredible year for their open-source Linux driver and how the open-source Radeon Gallium3D driver moved closer to performance parity with Catalyst. One of the lingering questions though is how does the Catalyst 14.12 Omega Linux driver from December compare to the latest Catalyst Windows driver? Here's some benchmarks looking at the latest open and closed-source drivers on Linux compared to the latest Catalyst Windows release.

http://www.phoronix.com/vr.php?view=21296

By far the best Phoronix test. I hope for Wine-Nine vs Windowz-D3D eventually and Heaven it's a good start.

[kalin]

He, he, i don't have 8.1, but i installed it maybe week ago on friend's laptop and run Valley benchmark there on Intel Pentium B950 with Intel HD graphics (which is awfull slow as it is OK), but i was surprised that OpenGL mode can't even start on that while DX9/DX11 mode works .

Together with these your result i somehow think 8.1 is not something reliablile to test OpenGL... i don't use Windows daily only for tests, but is Windows 8.1 realy that broken for OpenGL or am i miss something Or if someone has Window 7 vs Windows 8.1 OpenGL benchark results on same hardware, might be that will approve claim

8.1 is complete shit.I have one and bitterly regret. This is the worst spend money in my life and the reason to try linux as main os

[Sonadow]

8.1 is complete shit.I have one and bitterly regret. This is the worst spend money in my life and the reason to try linux as main os

I like 8.1 myself. Enough to actually upgrade my brother's and parents' computers to it.

[Kano]

Win 8.0/1 seems to have got some kind of weird self destroy function. It mainly happend with my Haswell box, not sure what caused it. I didn't reinstall it yet, but it happend with integrated SSDs and external via USB 3.0 connected HD. Win 7 SP 1 i dislike to install even more to play some games just because of the extreme amount of patches needed to download and install. I don't like to spend over 1 GB only for updates from my 40 GB highspeed traffic/month.

[drSeehas]

Which is fine, that is all what we want

Oh, you speak for all of us? I didn't know :-(
Why all the work for Wine and/or nine?

If Microsoft and AMD wants to put those on us, that should be fine... also PS4's API if Sony wants to put that on us, that should be fine too

Microsoft won't, that is for sure, but you ignore Wine and nine.
AMD Mantle: Hello bridgman.