AMD Sends In 2nd Round Of AMDGPU Radeon Driver Updates For Linux 5.3 - No Navi Yet

[dwagner]

... it could be any other piece of software or hardware that is influenced by dc=0. Maybe you have a faulty RX480 for example...

Even if there was such a highly improbable "subtle" hardware defect that was causing me (and many other people) to experience this kind of crashes (but only with newer Linux kernels, not with older, and - according to people dual booting their system also not under Windows), then the mere possibility of such a freak-bug (that no dealer will provide a refund on) would be reason enough to not consider an AMD GPU for my next system.

[lucrus]

so I think it's pretty clearly an amdgpu problem.

You experience crashes, I do not. We both have been using AMDGPU with a number of kernels since 4.13 up until 5.x. The only clear thing is that there must be "something else" that you use and I do not use (which pretty much excludes AMDGPU) that renders your system unreliable. I've suggested it was Ubuntu, but it could be any other piece of software or hardware that is influenced by dc=0. Maybe you have a faulty RX480 for example...

[Spazturtle]

This assumption smacks squarely in the face of Su who said it's a new architecture from the ground up. Rumors get old real fast. That one has spread for weeks on end and even resurfaces now after she was adamant that it's a new architecture.

It's a new microarchitecture (what most people mean when they say architecture), but it is still GCN ISA.

[dwagner]

I can't tell for sure, but if I were to bet on it, I'd bet this is a Ubuntu problem, not a AMDGPU one.

Well, I'm experiencing the instability under Arch Linux, with both the Arch-provided "stable" kernels and with kernels I compiled from amd-staging-drm-next myself. Also, reverting to very old kernels (4.13) allows for much longer uptimes between crashes than younger kernels, as does dc=0 - so I think it's pretty clearly an amdgpu problem.

[Marc Driftmeyer]

Assuming RDNA is strongly based on GCN, it's not like these updates aren't going to affect Navi.
Haha also, Phoronix has often been one of the leading sources to leak GPU information thanks to open source drivers. AMD may have caught on about this and told the devs to keep Navi out.

This assumption smacks squarely in the face of Su who said it's a new architecture from the ground up. Rumors get old real fast. That one has spread for weeks on end and even resurfaces now after she was adamant that it's a new architecture.

[schmidtbag]

Assuming RDNA is strongly based on GCN, it's not like these updates aren't going to affect Navi.
Haha also, Phoronix has often been one of the leading sources to leak GPU information thanks to open source drivers. AMD may have caught on about this and told the devs to keep Navi out.

[lucrus]

The only "big change" I have been waiting for the last 2 years was a driver not crashing. Alas, that did not materialize.

A colleague of mine who had previously claimed his RX480 was running fine with amdgpu just recently joined the club of the disillusioned when he upgraded to a new major Ubuntu release, now coming with a kernel that uses amdgpu.dc=1 by default. Now his previously stable system is just as unstable as mine.

I can't tell for sure, but if I were to bet on it, I'd bet this is a Ubuntu problem, not a AMDGPU one. I've been using a RX580 for quite a while now, and it always worked flawlessy on Debian with different kernels, some of which were vanilla kernels I compiled myself, like the current 5.2rc1 I'm using just now.

[Termy]

The only "big change" I have been waiting for the last 2 years was a driver not crashing. Alas, that did not materialize.

A colleague of mine who had previously claimed his RX480 was running fine with amdgpu just recently joined the club of the disillusioned when he upgraded to a new major Ubuntu release, now coming with a kernel that uses amdgpu.dc=1 by default. Now his previously stable system is just as unstable as mine.

interesting, last time i had driver issues with AMD was when catalyst was still de facto mandatory, my rx580 worked flawless without and with DC. Maybe Ubuntu-Specific?

[Peter Fodrek]

I think Navi is so secret that AMD is unable to release any RDNA code until AMD Next Horizon Gaming at E3

June 10, 2019 3PM PDT
The Novo | Los Angeles

https://www.amd.com/en/events/e3

[wizard69]

The only "big change" I have been waiting for the last 2 years was a driver not crashing. Alas, that did not materialize.

A colleague of mine who had previously claimed his RX480 was running fine with amdgpu just recently joined the club of the disillusioned when he upgraded to a new major Ubuntu release, now coming with a kernel that uses amdgpu.dc=1 by default. Now his previously stable system is just as unstable as mine.

This is why I really don't care about instant Navi drivers. I'd rather wait for stable drivers or at least to shake out the glaring bugs.