AMD Improving Linux Experience When Running New GPUs Without Proper Driver Support

An anonymous reader shares a report: While AMD provided upstream open-source driver support for the Radeon RX 7900 series launch, the initial user experience can be less than desirable if running a new Radeon GPU but initially running an out-of-date kernel or lacking the necessary firmware support. With a new patch series posted AMD is looking to improve the experience by being able to more easily fallback to the firmware frame-buffer when their AMDGPU kernel graphics driver fails to properly load.

With the new IP-based discovery "block by block" approach to how the open-source AMD Radeon Linux graphics driver is managing the hardware initialization with RDNA3 and moving forward, the AMDGPU driver will try to probe all Radeon GPUs even if it might not end up being fully supported. In turn that ends up destroying the system firmware frame-buffer. But right now in the case of booting an RDNA3 GPU with a slightly out of date kernel (pre-6.0) or lacking the necessary RDNA3 firmware for hardware initialization, it can mean the screen freezing or system appearing unresponsive.

Re:

[Narcocide]

Stale rant by several years; since switching from radeon to amdgpu, the Linux driver support for AMD cards is far superior to the equivalent Nvidia offerings.

Re:

[Shinobi]

Maybe to casual users. On the 3D content creation, 3D CAD, video post-prod as well as heavy compute loads, nVidia are still better, unfortunately.

Signed, someone who'd love to switch over full time to a Radeon Pro WX, or even just as the main card :/

Re:

[drinkypoo]

I'd prefer to have cards with OSS drivers, but what actually matters is having drivers that work.

CUDA remains the standard for GPGPU for a reason, and the reason is that as much as it sucks, it sucks less than the alternatives.

Before Stable Diffusion was released, I was considering AMD for my next GPU... so much for that idea.

Re:

[drinkypoo]

Found the Capitalist

That assumes that capitalism is the only way to get drivers that work. IOW, you're the capitalist.

Re:

[KiloByte]

The very use of the word "capitalism" suggests he's a commie. There's no such thing as capitalism, that's just a strawman concept invented by early tenebraries of worlds 2nd worst murderous ideology (only christianity being worse). It's a moniker applied to every single economy that 1. postdates the invention of money, 2. is not communist -- no matter whether its an early monarchy, mercantilist, free market, free market's antithesis ie bailout-everyone, feudal, dictatorial, etc, etc. I was teached at sch

Re:

[Narcocide]

Well, I guess I primarily meant from a stability and integration standpoint as well as a long-term support standpoint. NVidia cards still have some features the AMD cards don't (like CUDA), and still do a couple things in a way that looks better (like low-res hardware video decoding and upscaling) but their xrandr features implementation, Xorg integration and general distro package-management competency is complete trash and the drivers regularly suffer severe performance and stability regressions the likes

Re:

[Shinobi]

Funny that you claim increased stability and integration, because one of the issues I have with my AMD card is that when it crashes, it takes the entire system with it, while if the nVidia card crashes, most of the time I just need to restart either the driver or X.

As for installation, I use Gentoo and Funtoo, and last I had troubles with that was back in... 2007 or 2008, which is 14-15 years ago, so I do believe a lot of the issues come down to how distros handle things.

As for removing support for old card

Re:

[DamnOregonian]

Bullshit.
I use both.
They're both fine.

Re:

[KiloByte]

the Linux driver support for AMD cards is far superior to the equivalent Nvidia offerings

That's not a very high praise, you see...

Open source?

[antdude]

Didn't AMD open source its Linux drivers? If so, then why is its support so bad? :(

Re:Open source?

[drinkypoo]

Yes, they did. AMD historically had very bad drivers on all platforms, so it's not a surprise that there are still problems. The quality of the drivers has improved immensely since they made this change. Arguably, in most regards they are now superior to the nvidia drivers.

Re:

[antdude]

The last time I used their stuff was for ATI Radeon 4870 (512 MB of VRAM) video card in updated Windows XP Pro. and 64-bit 7 HPE. Before that, It still works too even though its drivers had minor issues. Before that, ATI Radeon 9800 Pro AIW (128 MB) in both updated Windows XP Pro. and Linux. Linux drivers were annoying to set up compared to NVIDIA's.

Re:Open source?

[drinkypoo]

The Linux drivers are generally excellent today, they do sometimes lag slightly behind new hardware but rarely for long. Meanwhile the Windows AMD drivers are still crap. I had crashes in the video driver on my Ryzen 3 laptop, works great under Linux. It has pathetically few cores, but I knew what I was buying (and it was seriously cheap, and the CPU is seriously fast for being cheap.)

I am using nvidia now and I'm glad because I'm using Stable Diffusion, and that's more problematic on AMD. But for other purposes I'd rather have AMD.

Re:

[vyvepe]

AMD Radeon VII driver sucks both on linux and windows :)

Re:

[bill_mcgonigle]

> Didn't AMD open source its Linux drivers?

Only a tiny part. I wound up doing a Ryzen 5600G for transcoding on a Jellyfin server. It's very energy efficient, but binary blobs are all over the damn place.

It mostly works, though, so that's a plus. Something is still wrong with MPEG-2 at low resolution; some people are starting to look into it.

Still, that wasn't in the test suite of a graphics company? Why does ffmpeg do a better job that AMD? What the hell are they spending their money on?

Re:

[F.Ultra]

They did and their support isn't bad, TFA is about running on a GPU that requires newer drivers than what you currently have installed but still make the GPU work.