Nouveau's OpenGL Performance Approaches The NVIDIA Driver

[AlbertP]

Strange that other people have so much problems with this chip, while it works fine at me, with all distros I have tried. Or are the 8400M GS and the 8400 GS very different?

[Geri]

Of course, opensource 3d drivers from the community for nvidia gpus are a very nice work. But comparisons about nvidia official driver VS. opensource community nvidia drivers posted on this site is fake and lie.

Results are deliberately distorted. To compare the drivers, the article writers searched games where the opensource drivers are not blooding out so mutch (rare games, mostly with very old rendering pipelines). In reality, official nvidia drivers are 10-30x faster than the opensource nv drivers, just go and try a more realible software with for example doom3engine: like prey the last call, chronicles of riddick, doom3 itself, or try different new graphics engines like maker3d, or unigine, and see the results with your own eyes. DO NOT BELIVE this fake results you have readed in these articles.

[AlbertP]

You say that they are faster than nv. But nv is not Nouveau. nv is worse.

Nouveau is not bad at all. Visit a random Linux forum and you will see many threads about people having problems installing the NVidia driver. I have helped some people installing Nouveau 3D on the Linux Mint forum, for using Compiz, and they're happy with it.

It lacks auto-detect of 2nd screens on my 8400M GS. That's the main reason why I don't use the binary driver anymore. And, I had to reinstall it every 2 weeks to keep my webcam working. 'Failed to capture image' showed up if I did not reinstall the graphics driver every 2 weeks. Compiz, KWin and my 2nd screen work so for me Nouveau is the best driver.

And why on earth should you need framerates of more than your screen's refresh rate?

[BlackStar]

It lacks auto-detect of 2nd screens on my 8400M GS. That's the main reason why I don't use the binary driver anymore.

Seconded.

And why on earth should you need framerates of more than your screen's refresh rate?

E-penis enlargement, mainly.

Back in the olden days, the drivers didn't even offer an option to disable vsync. Then benchmarks came and disabling vsync was the easiest way to make themselves look better than the competition.

Benchmarks should answer one simple question: given a specific game, what is the highest resolution that runs full speed with all display options enabled (i.e. 60fps for >80% of time)? Everything else is fluff.

[AlbertP]

What I meant to say is that there's also a side where Nouveau beats NVidia.

The driver is very reliable, if it works once, it also works a second time. This can't be said from NVidia: it often fails to install, on the same system where it once installed successfully. And it can suddenly stop working OK. If you don't need advanced 3D or very heavy games, then Nouveau is often the best way (if your distro includes 3D support like Fedora 15 or an easy installer for Nouveau 3D like Linux Mint 11).

It is not true that the binary driver is always better. That's what I want to make clear. I admit that it has some good things, and for some people it's the best solution, but the blob being closed-source is not the only reason you could have for using Nouveau.

[Geri]

of course i have spoken about nv as ::nVidia the company::, i shorted it as ,,nv''. and not from the old nvidia module, sry.

so i meant this:
,,opensource nv'' = Nouveau
,,official'' = official nvidia linux driver

for example, here is this: http://Maker3D.tk
It runs with 35-100 fps on official drivers on a nv260gtx. I not yet tried it with the opensource driver, but i have a guess: 3-20fps.

For this application, nobody created any application specific optimisation. Not nvidia and not in nouveau.

[Geri]

i have logged in to nouveau official irc channel and somebody checked the thing on a eforce 8400M GS, it was able to produce 7 fps with nouveau. Better than excepted.

[Geri]

.... original nvidia driver was able to run with 9 fps. however, nouveau lacks the AA.

[AnonymousCoward]

Seconded.



E-penis enlargement, mainly.

Back in the olden days, the drivers didn't even offer an option to disable vsync. Then benchmarks came and disabling vsync was the easiest way to make themselves look better than the competition.

Benchmarks should answer one simple question: given a specific game, what is the highest resolution that runs full speed with all display options enabled (i.e. 60fps for >80% of time)? Everything else is fluff.

Yeah, but VSync can result in poorer performance if triple buffering isn't used. Not always, but in at least some cases. So it's understandable that some don't use VSync if they are not hitting the 60fps target with it.

Additionally, older CRT monitors often hit refresh rates up to 100hz or so, so one would gain benefits from hitting 100fps.

Finally, benchmarks that hit above 60fps can give you idea of how much a card will scale in the future, and how it compares to performance of other cards.

[AlbertP]

My CRT from 1996 supports up to 75Hz, and some may even support more as AnonymousCoward suggested, but they won't go up to 200Hz so the framerates in the test can't be reached on such a screen.

Anyway, 60Hz (59,9 on my laptop) is a very common (maximum or fixed) rate for modern flatscreens. I think a framerate higher than that should just be considered '60'. They should compare visible performance and no invisible >100Hz rates.