Some AMD Grenada Cards Have Been Borked On The Open-Source Driver For 2 Years

[Masush5]

It seems Phoronix launched some anti AMD strategy last days. It is known for propaganda articles.

Are you serious about that? To me phoronix feels like the biggest gathering of amd fanboys to be found on the internet. Probably ~70% of Micheal's recent benchmark articles are about AMD hardware on OSS drivers. But you think he launches a hate campain because he points out issues with a specific line of cards and chooses nvidia for his workstation?

btw: I think most of the fanboism is deserved I exclusively own amd gpus.

[R41N3R]

Well if a bug report is open too long, I can understand some frustration... but I don't see that these discussion based on the news are constructive or even helpful at all, instead I think it leaves wrong impressions. As a user of these cards I would have Radeon and AMDGPU driver to play around to see what works best, I would call this quite comfortable ☺️

[Guest]

I can confirm - I have an XFX Radeon R9 390, lots of problems with the Radeon driver - AMDGPU is a lot more stable, and seems to perform better. The only downside is lack of HDMI audio (for now).

There was one annoying bug with 4.12, random crashes would occur (crash = blank screen and can't switch to TTY). Compiled 4.13rc1 and it works fine now. Also, I'm super happy about the new module params, don't have to compile the kernel anymore to use AMDGPU.

[Create4Life]

I have a vtx3d r9 390 8GB.
Ever since fglrx has been deprecated I have been left without any supported driver.
For the intent of running linux this GPU has become a very expensive paperweight and has been the source for a lot of frustration.

Without forcing "radeon.dpm=0" I can neither boot the radeon backend nor amdgpu which immediately blackscreen and completely hang the whole os a second after loading any gui.
With this "workaround" added I downgrade my last gen high end grafics card to slower than an entry level APU so I do not consider this to be a working workaround. The closed source Amdgpu pro even though technically supported suffers from extreme display corruption and instability (even in single user command line mode). I dont think I ever managed to open an application before the system crashed.

I have also been left out of using opencl, which was the main reason I even built this workstation since neither amdgpu-pro nor rocm work.

I am grateful to AMD for pushing open source but they completely abandoned everybody relying on their closed source driver which was finnicky but at least was a workable solution to many of their clients.

By this time I probably should have returned the card but as it was running perfectly fine on catalyst and windows I was hoping the "transition period" to the all open stack would not drag on and drag on for as long as it did and still does. If I recall correctly it has been more than a year that I have been unable to do any work on linux and have to fall back to windows to do allmost anything.

At the current pace my hawaii chip is going to be completely outdated before this regression is fixed. The fact that it used to work perfectly fine using their old proprietary driver makes all of this even worse to me.

[Guest]

Well I just compiled 4.13rc1 yesterday night, will see if I run into any issues. I did encounter the black screen problem, but only recently, and it seemed to happen quite frequently while on 4.12 stable release.

[bridgman]

The closed source Amdgpu pro even though technically supported suffers from extreme display corruption and instability (even in single user command line mode). I dont think I ever managed to open an application before the system crashed.

I don't remember hearing about this before - just checking, is this with the AMDGPU-PRO driver installed on a stock supported distro (ie without a new kernel) ? I'm asking because there was some bad advice going around saying that a custom-built kernel was required to use AMDGPU-PRO on CI, and that caused a lot of grief for customers.

[leipero]

Michael had a 290 acting up which is basically the same card. The issue was never widespread, which made it difficult to pinpoint and fix (I'm not sure if it was ever pinpointed or just fixed as part of larger changes).
So yes, I expect "but this works on my machine" experience to be much more common. Yet, there's a chance some quirk of this silicon just likes to rear its head now and then.

I really doubt it's chip problem, what would be useful is, for every user to list it's configuration, and if GPU have same problems on different platforms (I assume Michael already tested that first, and it does have problems) naming manufacturers and models of GPU's would be useful. Back in the days of FX GPU's, i had two identical low end FX-5200 GPU's, one from MSI (64-bit version) and one from Gigabyte (128-bit version), and I'm talking about Windows here, with same exact driver, same clocks, you would assume 128-bit version would outpreform 64-bit one, but that wasn't the case, I had serious problems in Colin Mcrae Rally 3 at the time with that GPU, while MSI did work great. That might be the case here, BIOS or design problem, not chip problem.

[Apopas]

It seems Phoronix launched some anti AMD strategy last days. It is known for propaganda articles.

Michael should add a Dislike button as well, next to the Like one.

[pal666]

Probably ~70% of Micheal's recent benchmark articles are about AMD hardware on OSS drivers.

well, neither nouveau nor intel are fun for benchmarking, so 70% is pretty low

[Guest]

Are you serious about that? To me phoronix feels like the biggest gathering of amd fanboys to be found on the internet. Probably ~70% of Micheal's recent benchmark articles are about AMD hardware on OSS drivers. But you think he launches a hate campain because he points out issues with a specific line of cards and chooses nvidia for his workstation?

btw: I think most of the fanboism is deserved I exclusively own amd gpus.

It's also one of the biggest gathering of Linux fanboys. However, it didn't stop him from attacking Linux and GPL in the past.