22-Way Comparison Of NVIDIA & AMD Graphics Cards On SteamOS For Steam Linux Gaming

[Passso]

It would be better to see tests on some i7 6700k or 4790k OCed to 4.5GHz or even i5 6600k (also 4.5GHz) as they are what typical gamer has, a

oO No, this is definitly not what gamers have, this is what US gamers may have.

In the world you will get far more i3, old i5 and even Pentium.

[Passso]

The open-source drivers (r600g included) can't even run all the tests done here.

Please Michael, save your precious time and do not answer. Duby229 and dungeon are just repeating the same story concerning AMD and OSS drivers from ages like sick Alzheimers
But you know what, I love them anyway because even when they agree they are still fighting together!

[Passso]

I don't have enough fundamental understanding to say. I don't think anybody really knows. I've read arguments that claim there exists a hardware based instruction scheduler that is to blame, but of course now I can't find that link. It was on mesa's mailing list that I read that. So really I don't know.

EDIT: I don't know if caring about OpenGL is the answer, but I suspect that has a lot to do with design decisions made during hardware development. Nvidia cared and AMD didn't it seems is the best answer I can make. I know it's not a technical answer, but it seems logical that has unknown technical repercussions.

So you are telling us that AMD do not care about OpenGL since conception
=> So they do not care about open and Linux graphical libraries.
=> So they do not care about Linux gamers

BUT your conclusion is that we must buy an AMD card.

You should go to a vegetarian forum and prove them that beef has more C vitamin than oranges, you're good enough!

[drSeehas]

The entire GCN series only started two years ago, ...

???
GCN started January 2012.

[duby229]

So you are telling us that AMD do not care about OpenGL since conception
=> So they do not care about open and Linux graphical libraries.
=> So they do not care about Linux gamers

BUT your conclusion is that we must buy an AMD card.

You should go to a vegetarian forum and prove them that beef has more C vitamin than oranges, you're good enough!

That's not at all what I said. My very own words are in your quote, please read them.

It's pretty damn obvious AMD or ATi before them didn't care enough about OpenGL to make sure it performed well.

EDIT: I suppose I must not forget that when other people don't have a good argument they make shit up and then try to pretend like it was you that did it. Nonsense. Stupid retarded nonsense. Your words are your words and not mine. This forum has a working quoting function, please use that to quote me instead of making shit up. I never said a single word about vegetarians or oranges only you did.

[Taupi]

Thank Michael L. for this very interesting study. As usual, nice work and keep it up!

Is there some place(s) where we can tickle AMD so they put some more forces on linux drivers? Petition, kickstarter to have better open-source drivers...? ^^
Will Vulkan change the situation (and can we hope more about)?
... I'm still awaiting for GOG-Galaxy for linux to get rid of dual boot!
I have a R290 and I should have bought a Nvidia 750 to save my money!

[drSeehas]

... for those that need ... features still lacking in the open source ones. ...

And why doesn't AMD add the missing features to the open source ones instead of the proprietary one?

... i wanted to see here ... APUs also tested...

+1

[thelongdivider]

yep, after spending years with a notebook with a 4850 and the shitty linux drivers with slightly better windoze drivers I vowed NEVER ATI again. The 4850 notebook was a result of my giving them one last chance, although I did build an entire AMD 'nostalgia' system fx-9590, r9 280x, and did pick up an a10-7850k to experiment with but as usual AMD software lags generations behind their hardware and they're no better than ati at software.

Additionally since ATI branched out into heating augmentation I'm even happier to have dropped them as a potential supplier. I'm not even convinced that GCN is a good arch, and am skeptical about HBM but that may just be more ATI sw shittiness.

[EDIT]
Catalyst is really just a corollary of having to produce a driver for the more important commercial/pro cards, the fire series. Really the same thing for nvidia but they appear to actually take their sw seriously.

As to ATI drivers better since 2009?! GTFO man! They're just as shittily broken now as they were 9y (and more) ago. Hell I've had instances where their shitty released driver wouldn't even completely install, failing randomly in thee process -> rolling back to the old 'stable' driver. The only improvement that I can note is that they finally fscking figured out how to restore the display in windows AFTER replacing the video driver whereas before you got a black screen, and got to guess if it was finished installing under windows, meanwhile as far as I can remember nvidia ALWAYS managed that.
[/EDIT]

This is my point. Catalyst is unusable from a desktop perspective on linux (get modern X.org support, or better yet wayland). RadeonSI is excellent, but using my dual card system with those drivers seems like a waste.

[darkbasic]

That particular test results is fake , no one else had not reproduced this kind of performance . Even author commented on irc i think, that something getting wrong there.

The test is not fake, simply the merit was not of the si scheduler but instead of a commit which got reverted between the two FOSS tests.
I had been able reproduce the si scheduler numbers with git master and another user reported he got similar results too. I don't know if foss numbers regressed in the meantime, but this is what you got at the time.

[dungeon]

The test is not fake, simply the merit was not of the si scheduler but instead of a commit which got reverted between the two FOSS tests.
I had been able reproduce the si scheduler numbers with git master and another user reported he got similar results too. I don't know if foss numbers regressed in the meantime, but this is what you got at the time.

I know i bisected that regression maybe few days before you made that results, so i knew it is wrong on arrival with that sched improvments on top of the bug.

OK it was not fake, but (at least to me) expectedly wrong because i knew for days before that there is perf regression flawing around.