Open-Source "Nouveau" Driver Now Supports NVIDIA Ampere - But Without 3D Acceleration

[oiaohm]

I have a 1080ti for 3 and a half years with proprietary drivers and have had zero problems with the card, you can blame Ubuntu for that one. Also on my end there aren't any hacks, running Manjaro which is just wrapper over Archlinux.

Next time when you complain about something, choose the right target.

Not exactly that some usage issues that come up. Like if you are switching between text mode and graphical a lot that Nvidia binary driver starts causing system wide lockups this is not a distribution thing this is defect in way Nvidia driver hooks into the system.

[undefined]

As long as NVIDIA's proprietary driver supports it, what's the problem? All the mainstream distros have easy installs for NVIDIA's driver.

And when it doesn't? If in 2 years Nvidia writes on their webpage that "we, gamers, and the rest of the industry do not use Linux and GPUs without SuperRTX3.0 anymore, so we are dropping Linux support", what are you going to do? Disable updates of your OS? Practically, your GPU will become a piece of garbage. You won't even theoretically be able to do anything with it without their blessing. It is almost like you don't even own the GPU, you need a blessing from Nvidia to use it. And God forbid you disagree with their EULA.

[mdedetrich]

I really wonder from what standpoint all the nvidia fanboys here come from.
I have a 1060, a VII and a RX 560 in my at work workstation. Imagine which of those cards is nothing but pain.
Showing a friend of mine the joy of linux, the only problem was him having a 1080 Ti, it resulted in nothing but pain and that with Ubuntu, the supposed nvidia out of the box distro.
Putting that damn RX 560 in that workstation was a life changer because it f*ing works, I had not to bother about the GPU once in a year of tracking the newest kernel releases and getting some work done.

And just thinking about the amount of code every single desktop on our platform that had to incorporate hacky tricks to get the desktop working properly on Nvidias own driver. No matter if you use Gnome, KDE, XFCE, Mate. This affects us all.

Edit: This is not even about Nvidia vs AMD. No matter if you have a Intel, AMD, Mali, Adreno, VideoCore and many more, it just works. The only two odd ones out are Nvidia and PowerVR.

I have a 1080ti for 3 and a half years with proprietary drivers and have had zero problems with the card, you can blame Ubuntu for that one. Also on my end there aren't any hacks, running Manjaro which is just wrapper over Archlinux.

Next time when you complain about something, choose the right target.

[torsionbar28]

I don't get why amd users always are writing toxic posts in nvidia related news

Because so many amd users are former nvidia users, having abandoned nvidia due to their anti-FOSS stance and the endless frustration caused by their proprietary driver. I'd be mad too if I had to deal with all that. Try having some empathy for the hours of frustration and lost productivity others have endured at the hands of nvidia.

[mangeek]

looks like I get to miss out on the fun of the software-accelerated cursor and losing everything when the DE decides to crash (GNOME).
'm not convinced any other DE or WM is usable in Wayland to bother with, yet anyway. But guess what? Xorg works great still

I've been daily-driving GNOME on Wayland for eight months, on a laptop that I plug and unplug from a dock to a HiDPI screen that needs fractional scaling. It works beautifully. Not a single DE crash.

You complain about unaccelerated cursors, but the awful performance of fractional scaling on X is what got me to switch to Wayland.

[dweigert]

All of my computers use Nvidia cards, I'm running OpenSuse Tumbleweed on most of them. I don't have too many issues with the proprietary drivers, just the occasional wait for them to release the next one as they work to accommodate the kernel changes. in my two servers (Dual intel Xeon E5-2673 v4 CPUs and 256 GB RAM in each) , I run a 1080Ti and a TitanV. In my workstation (AMD Threadripper 3970 CPU and 128 GB RAM), I run a 2080Ti, but will be changing that out for a 3090. The 2080Ti will probably replace the 1080Ti, but that isn't a given. Please note, that these three machines have NEVER had a driver issue that causes a hang. Each of them is used for testing things like database clustering and virtual machines, and containers, with side forays into AI processing.

I get that a lot of people don't like the proprietary drivers. Trying to unencumber them from things licensed from SGI and other companies is a nightmare, simply because there are lots of lawyers involved. Some of those companies are defunct, but contract law holds them to the license agreements. I support the work that Nouveau does, but I'm going to use the best hardware for the job that I need doing.

[qarium]

The last AMD/ATI card I used was a Radeon 9500 Pro and I will freely admit it was by far the dominant card at the time and an amazing feat of engineering! I really wish I could see another AMD card achieve what the 9XXX series did. At the same time 90% of my computer work these days requires CUDA so until ROCm becomes something real AMD cant touch Nvidia.

https://geizhals.de/?cmp=2448120&cmp=2366527

6900XT=1449€
3090XT=1696€
you save 247€

isn't a 6900XT a technical master peace?
as soon as the DLSS2 feature is done as AMD Super-Resolution on AMDs Driver you can be sure the 6900XT is always faster.

why i am sure about this? over all benchmarks i have seen the 3090 looks faster at 4K and the amd card is faster in 2K and 2,5€ and also 3K. as soon as you use DLSS2/AMD Super-Resolution you run the game at a lower resolution and upscale it to 4K but this means as soon as AMD Super Resolution is done in driver there is a high chance that the 6900XT always wins. also all only watch nvidia implementation of Ray-Racing but if you watch the games who use the AMD implementation then the 6900XT is 10% faster than the 3090. sure the Nvidia implementation is faster on the 3090. but because the new Xbox and Playstation is AMD based you will see the AMD implementation in many games.

about CUDA vs ROCm yes you are right if you need cuda you maybe choose a 3090 anyway.
but i see the future of Compute in the Vulkan based Compute. as soon as this happens the 6900XT will shine.

[oiaohm]

I had 5 different Polaris GPUs (2 SAPPHIRE RX 580s, 2 XFX RX 580s, and 1 XFX RX 560); all of them had instability at 4K@60Hz over HDMI, and it happened on Windows, macOS, and Linux, on several different motherboards, about 6 HDMI 2.0-certified cables, and even a TB2 eGPU enclosure. There's other reports about this from others on AMD's forum and Hard Forum.

After the 3rd GPU, I was questioning if I was really that unlucky, but after trying a GTX 1060, I had no issues whatsoever. But then silly me figured maybe it was just XFX being odd and I decided to give SAPPHIRE a try. Both SAPPHIRE cards had the same instability, and still have it today.

With the Polaris I have SAPPHIRE RX570 that was got in the last 12 months done with the new process of production about 2 degrees hotter in the reflow it is in fact happy with the 4k@60Hz. I also have access to 8 SAPPHIRE RX 580 that have been repaired by a taking the main chip of and reball it and putting it back on because the had completely stopped and they don't have the 4K@60hz problem please note the documentation on some of these before end up for recycling say they were not doing 4K@60hz before having the major operation. GTX 1060 from 2016 depends on vendor and batch some of those have the same construction problem with the reflow temperatures and the same faults so you could have been lucky with the GTX 1060 not all who bought a GTX 1060 were lucky enough to get working 4K@60Hz. So maybe by the time you got around to getting a GTX1060 you got a 2018 late 2017 one? Even new cards you can get unlucky in the department.

The ones where the card due to the bad solder just complete stops have had a far better time of it not spending ages in diagnostics really not leading that far. Yes it lead to a lot of false starts where person will think windows or linux is working until the run long enough and the one they think works bits it when power management goes the right/wrong way and it fails.

Espionage724 there is such thing as bad timing and bad lucky. The fact the cards are still working just without 4k@60 you have been luckier than some. Those 8 that were reworked were all 100 percent zero output before being reworked. Please note there is also a mountain of Nvidia gpu from the same time-frame also needing rework so they will fully function as well..

[Danielsan]

There is absolutely nothing in the Linux community that sets the world on fire like Nvidia news...

It is pretty nasty that Nvidia won't release these PMU signatures to extend the lifecycle of their products as well to allow to use opensource driver fully. It doesn't make sense for Linux which is a ridiculous small portion of the desktop market, it won't affect Nvidia business while will help in its reputation, it is just nasty.. Having full access to their old GPUs with free software can have also meaning in the research field, will benefit a lot of other areas while for instance will prevent many to continue to purchase their products...

[Sonadow]

Nvidia on Linux is not the best choice if you don't want to spend the days between black screens and various problems and it is enough to open a forum of any distribution to understand that it is a problem .

Speaking as a Nouveau user on Gnome Wayland and a GT1030:

That's utter BS.

After that if a user wants Nvidia hardware it is a choice, of him but don't complain if things don't work as they should.

I get a much sharper and clearer display on Nvidia hardware compared to AMD hardware on the same monitor. That makes AMD hardware trash-tier from the beginning. There is no point prostituting the "awesomeness" of the open drivers for AMD when the output quality from the card is already subpar.