XFX RX590 resets computer in games on Ubuntu 20.04, help needed...

[zzarko]

I have XFX RX590 Fatboy 8GB, I5-2500K, 8GB RAM, Ubuntu 18.04 with 5.6 kernel. In two Steam games (so far), Black Mesa (native Linux) and Sherlock Holmes Crimes and Punishments (via Proton, any version), when I get to a certain spot (always the same spot in the game), the computer just resets. I have checked the temperatures of the CPU and GPU (by writing into file every 2 seconds), they are way below threshold (~40C for CPU and ~60C for GPU), and syslog, kernel.log and Xorg.0.log show absolutely nothing abnormal when this happens (in all of them the reporting just stops when this happens, there is no any trace of what could be happening).

Today I have installed Ubuntu 20.04, and with the default configuration, the computer resets on those two games happen at the exactly same spots. I have tried Oibaf PPA with 20.04, but that made computer unusable as I had desktop freezes and garbage right after logging in, so the packages were removed.

Next, I have tried to add amdgpu.drm=0 to the boot line, and that solved the crashing, at the cost of performance. I have also tried to put 'high' in power_dpm_force_performance_level, that causes a computer to freeze instead of resetting when game gets to that specific point.

The card is under warranty, but I'm not sure if it is hardware or driver problem. Do someone has any advice? Is there something I could also try?

[Anarchy]

If you get crashes consistently at those two games at some specific place in the games you could try enabling netconsole and just log the messages on a different computer or check how to properly configure and enable kdump, you can also enable oopses as panics. in certain cases it can produce a kernel dump that can be later analysed to figure out what went wrong.

I have Black Mesa & ubuntu 20.04 & an rx 580. where does it crash for you? I finished the game fine on my machine, but I could check again.

[zzarko]

I have Black Mesa & ubuntu 20.04 & an rx 580. where does it crash for you?

Right at the introduction (before any actual gameplay), after loading second part of train journey, when train stops and begins to descend vertically.

[chithanh]

I would guess faulty PSU.

Maybe try limiting the maximum power which the card can consume via /sys/class/drm/card0/device/hwmon/hwmon0/power1_cap

[Anarchy]

Right at the introduction (before any actual gameplay), after loading second part of train journey, when train stops and begins to descend vertically.

Hmmm... I remember having some problems around there because mesa was too old in 18.04. Gonna check it and let you know soon. Is it possible for you to install any of the dev mesa ppas on ubuntu & test again?

[zzarko]

Hmmm... I remember having some problems around there because mesa was too old in 18.04. Gonna check it and let you know soon. Is it possible for you to install any of the dev mesa ppas on ubuntu & test again?

I have tried padoka ppa, with the same results, but I'm willing to try others. Also, on 20.04 (mesa 20.0 if I remember correctly), I have the same problem. I'll try to find how to use kdump and/or netconsole. What is strange is that there is no trace of anything in the logs... I would expect that maybe xorg log has something before restarting the computer, but no... I'll also ask around to borrow some other, more powerful power supply (although, my previous card was 7950, and this power supply was enough).

[Anarchy]

I have tried padoka ppa, with the same results, but I'm willing to try others. Also, on 20.04 (mesa 20.0 if I remember correctly), I have the same problem. I'll try to find how to use kdump and/or netconsole. What is strange is that there is no trace of anything in the logs... I would expect that maybe xorg log has something before restarting the computer, but no... I'll also ask around to borrow some other, more powerful power supply (although, my previous card was 7950, and this power supply was enough).

Yo, I checked again on my vanilla kubuntu 20.04, and it ran fine on default settings. I played it from the beginning to the big boom in the lab and on the xen levels. Everything was buttery smooth, no artifacts, no crashes, no stuttering, nothing. Although, when I later checked I was actually running the beta-version; I got Black Mesa when it had just come out, so the beta ran better on Linux. The only extra thing that I installed was libtcmalloc-minimal4, but that's because the Interloper level would crash on me right at the start.

You could try doing the same. Install the package above and try playing on the (default) settings the game picks for you. If you play it on 18.04 you need newer drivers than what the system has by default.

A few weeks ago, btw, I had gotten a 5700XT, I'm not sure if the card was a turd or the game is finicky, but I experienced a lot of crashes, black screens, various artifacts and system lockups when I tried to play around with the settings for too much. This is a well documented problem with Black Mesa on Linux.

[zzarko]

I remembered that I had an old win 7 install somewhere, and I have installed Black Mesa on it (with countless restarts for this or that for the updates and game installation). The game crashed same as in Linux.

I took me some time to get new power supply, CoolerMaster 700W. After the installation, win 7 version ran without problems. OK, not exactly, Radeon installer sneaked ZoneAlarm, which in turn decided that some game files have a virus and blocked the game from loading properly, resulting in a crash. As I do not use win anymore, it took me some time to find out what's wrong, as I forgot almost everything regarding win usage. I'm completely out of the world of custom installers for everything, antivirus software, and all that ...

Anyway, then I tried Ubuntu 18, game crashed and froze the PC (but, no reset this time). Ubuntu 20 ran the game just fine.

I wish to thank you all for the advises! It looks that XFX RX590 is more power hungry than my old card. My old power supply was a Chieftec 650W (I saw that it was 650W when I took it out, I was under the impression that it was 600W), I really thought that it would be enough, but maybe it degraded over the years of usage...