The NVIDIA GTX 900 Series Performance On The Open-Source Nouveau Driver

[Pecisk]

Well, additional firmware is a bit of setback, but according to devs it is highly likely they will get it. It will take time to get it working though.

This also clear some questions about me buying 9xx series - not gonna happen any time soon. Seems I will be picking up red for next upgrade.

[karolherbst]

Well, to add technical bits to the debate: in my case (660), the NVIDIA blob causes incredible amounts of tearing. As in, it's impossible to play anything 3D, even scrolling pages is really hard on the eyes. It's fixable with a whole bunch of hacks applied (including setting some environment variables; vsync is just a small part of what needs to be enabled), but it causes a huge performance penalty.

On the other hand, Nouveau works perfectly in that regard, there is no tearing whatsoever with default settings. However... it tends to randomly lock the GPU (and thus the system) up.

So it's a "pick your poison" situation for me. Sometimes I'm thinking that running swrast might be the best solution... At least it would not require rebuilding the blob every time the kernel updates (which, on Tumbleweed, is often).

you might want to report what dmesg throws out when the GPU locks up. We really want to get this stable and any bug report help here.

[chithanh]

you buy a 980 to run with open drivers? what wast of money

Maybe he is one of the many users who use Linux for work and dual-boot into Windows for games.

Power management would be good for that though, too bad that the article did not include any power measurements.

[darkcoder]

I will never use binary only blob crap. New kernel, xorg, etc? Wait for their binary only update. Want to do GPU driver and architecture research and development? Out of luck. Want to run on PowerPC, UltraSPARC or ARM64? Good luck either.

Binary only driver are cancer in the open source world.

I'm thankful to all the nouveau developers for being able to use 3D acceleration on my G5:

http://t2-project.org/hardware/workstation/Apple/G5/

Everyone who wants binary only blobs can go back to windows.

Huhhhh?


Last time I checked third party video cards relied on an X86 family CPU!

And kernel/xorg releases.. Yes that's a pain, sometimes to wait for support to be added. But those drivers are intended for already released distributions which BTW have a release and software version lock policy... Only bug fixing. If you use rolling distributions things may break by being first, and they (the rolling distros) already warn you about that.

But if someone like Ubuntu releases a new kernel package to fix a vulnerability, and leave the binary blobs out, then blame that distribution update policy. They should recompile any kernel dependant package and release the whole combo at the same time... Not a day or two later.

[magika]

you buy a 980 to run with open drivers? what wast of money

How in the world you've come to such conclusion? I'm fine with Ivybridge graphics when it comes to linux, but it's annoying to replug monitors each time I want to boot Windows to play them couple of games. Now that there is 3D with nouveau, I can use wayland, thats it.

[bridgman]

Huhhhh? Last time I checked third party video cards relied on an X86 family CPU!

The ATI/AMD open source drivers are used with a number of different CPU architectures - PowerPC, MIPS, ARM among others. Sounds like it's the same for Nouveau... might even be true for Intel but it would be hard to scrape the iGPU off the x86 die to find out

[bug77]

I will never use binary only blob crap. New kernel, xorg, etc? Wait for their binary only update. Want to do GPU driver and architecture research and development? Out of luck. Want to run on PowerPC, UltraSPARC or ARM64? Good luck either.

Binary only driver are cancer in the open source world.

I'm thankful to all the nouveau developers for being able to use 3D acceleration on my G5:

http://t2-project.org/hardware/workstation/Apple/G5/

Everyone who wants binary only blobs can go back to windows.

The thing is, the open source world is just a part of of the real world we live in. Probably the better part of the software world, but still, just a part.
Also, cancer? Really? I've been using nvidia's blob for 10 years or so and it hasn't spread to any other parts of the system yet.

[chithanh]

bug77 About the cancer thing: I don't know if you are old enough to remember this, but NVidia tried to spread proprietary drivers beyond just GPUs. The forcedeth driver for the Ethernet function of nforce chipsets was written as a free replacement for NVidia's proprietary network driver. NVidia soon lost interest in the blob.

The you have the situation with projects like Wine which are traditionally very much aligned with the NVidia graphics driver, sometimes to the exclusion of open source drivers (the rejection of gallium-nine patches comes to mind).

[rene]

The thing is, the open source world is just a part of of the real world we live in. Probably the better part of the software world, but still, just a part.
Also, cancer? Really? I've been using nvidia's blob for 10 years or so and it hasn't spread to any other parts of the system yet.

If more people would have boycott nvidia for the last decade we would have an even more health AMD/Intel open source ecosystem. And who knows maybe even Matrox or so would still be in the game (very open source friendly driver a very long time ago, ...)

[rene]

Huhhhh?


Last time I checked third party video cards relied on an X86 family CPU!

And kernel/xorg releases.. Yes that's a pain, ....

I do not know what you are mixing up there, with "third party video cards relied on X86 family CPU".
This sentence makes so no sense to me, ...

Last time I checked some old Radeon run just fine in my UltraSPARC, and the Matrox in the IBM RS6000 was working as fine as the Radeon or now Nouveau Nvidia in my PowerPC64 G5, ...

Forgot which card I had in the IA64 before that ship sunk, etc. pp: http://t2-project.org/architectures/

PS: Maybe your point was binary only crap was mostly available for iX86 only. Which would proof my point anyways, …