AMD RX590 locks up desktop under moderate to high usage

MT5496 · 10-10-2020, 03:01 PM

Hello, i am currently having an issue where if i load my GPU (AMD XFX RX590 Fatboy OC) with superposition. i am running 2 monitor both at 1440p and 144hz.
i am planing on upgrading to a new GPU when AMD announces their new lineup. But this issue is making me lose confidence in AMD for my next Linux upgrade.

This issue does not occur on Windows 10. This also does not occur with a NVIDIA GPU. (I get other issues with nvidia but that's not relevant to this error)

My Distro is Manjaro KDE and this issue seems not to be distro specific since it occurs on Linux Mint Cinnamon.
5.4.67-1-MANJARO
Here is the output of

Code:

sudo inxi -G -C -M -xxx

System:    Kernel: 5.4.6-1-MANJARO x86_64 bits: 64 compiler: N/A Desktop: KDE Plasma 5.19.5 Distro: Manjaro Linux 
Machine:   Type: Desktop Mobo: ASRock model: X570 Taichi serial: <filter> UEFI: American Megatrends v: P3.40 date: 08/11/2020 
Battery:   Device-1: hidpp_battery_0 model: Logitech M570 charge: 50% status: Discharging 
CPU:       Topology: 8-Core model: AMD Ryzen 7 3700X bits: 64 type: MT MCP arch: Zen 2 L2 cache: 4096 KiB 
           flags: avx avx2 lm nx pae sse sse2 sse3 sse4_1 sse4_2 sse4a ssse3 svm bogomips: 115236 
           Speed: 3599 MHz min/max: 2200/3600 MHz Core speeds (MHz): 1: 3599 2: 3599 3: 3600 4: 3598 5: 3599 6: 3600 7: 3599 
           8: 3600 9: 3599 10: 3600 11: 3599 12: 3599 13: 3599 14: 3598 15: 3599 16: 3599 
Graphics:  Device-1: Advanced Micro Devices [AMD/ATI] Ellesmere [Radeon RX 470/480/570/570X/580/580X/590] vendor: XFX Pine 
           driver: amdgpu v: kernel bus ID: 0e:00.0 
           Display: x11 server: X.Org 1.20.9 driver: amdgpu FAILED: ati unloaded: modesetting resolution: 1: 2560x1440~144Hz 
           2: 2560x1440~144Hz 
           OpenGL: renderer: Radeon RX 590 Series (POLARIS10 DRM 3.38.0 5.8.11-1-MANJARO LLVM 10.0.1) v: 4.6 Mesa 20.1.8 
           direct render: Yes 
Audio:     Device-1: AMD Ellesmere HDMI Audio [Radeon RX 470/480 / 570/580/590] vendor: XFX Pine driver: snd_hda_intel 
           v: kernel bus ID: 0e:00.1 
           Device-2: Advanced Micro Devices [AMD] Starship/Matisse HD Audio vendor: ASRock driver: snd_hda_intel v: kernel 
           bus ID: 10:00.4 
           Device-3: Cubeternet 4-Port USB 3.0 Hub type: USB driver: snd-usb-audio,uvcvideo bus ID: 2-4.4.4.4:5 
           Device-4: C-Media USB2.0 Hub type: USB driver: hid-generic,snd-usb-audio,usbhid bus ID: 1-5.4.4.2:7 
           Sound Server: ALSA v: k5.8.11-1-MANJARO 
Network:   Device-1: Mellanox MT26448 [ConnectX EN 10GigE PCIe 2.0 5GT/s] driver: mlx4_core v: 4.0-0 port: N/A bus ID: 04:00.0 
           IF: enp4s0 state: up speed: 10000 Mbps duplex: full mac: <filter> 
           Device-2: Intel Wi-Fi 6 AX200 driver: iwlwifi v: kernel bus ID: 07:00.0 
           IF: wlp7s0 state: down mac: <filter> 
           Device-3: Intel I211 Gigabit Network vendor: ASRock driver: igb v: 5.6.0-k port: f000 bus ID: 09:00.0 
           IF: enp9s0 state: up speed: 1000 Mbps duplex: full mac: <filter> 
Drives:    Local Storage: total: 2.05 TiB used: 528.44 GiB (25.2%) 
           ID-1: /dev/nvme0n1 vendor: Samsung model: SSD 970 EVO Plus 1TB size: 931.51 GiB 
           ID-2: /dev/sda vendor: Western Digital model: WD1003FZEX-00MK2A0 size: 931.51 GiB 
           ID-3: /dev/sdb vendor: Crucial model: CT250MX500SSD1 size: 232.89 GiB 
Partition: ID-1: / size: 916.10 GiB used: 181.07 GiB (19.8%) fs: ext4 dev: /dev/nvme0n1p2 
Swap:      Alert: No Swap data was found. 
Sensors:   System Temperatures: cpu: 36.6 C mobo: 48.5 C gpu: amdgpu temp: 68 C 
           Fan Speeds (RPM): fan-1: 0 fan-2: 2195 fan-3: 1335 fan-4: 1371 fan-5: 937 fan-6: 0 fan-7: 0 gpu: amdgpu fan: 395 
Info:      Processes: 357 Uptime: 11m Memory: 62.74 GiB used: 3.21 GiB (5.1%) Init: systemd Compilers: gcc: 10.2.0 
           clang: 10.0.1 Packages: 1594 Shell: Zsh v: 5.8 inxi: 3.1.05

Now the issue that is occurring is a lockup for my desktop, weirdly enough if im quick enough i cant get into a console with CTR+ALT+F2.
After some rainbow static graphics im met with a console, then after a minute or so both monitor go black and say no input.
my PC is still on and cannot recover until i do a hard reboot.

if i use corectrl to reduce the power limit of the GPU to 160 watts instead of 200 watts i can get through a superpostition run with no crash, but if i run a game for about 10 minutes i get a crash.
im not sure how to diagnose it since the logs ive found (Xorg.log.old, journalctl -b-1 -p err) doesnt show to much or isnt understandable to me.
i have attempted using a higher kernel 5.8 but issue is still there.

Xorg log: https://pastebin.com/ExrMCwjG
journal -b-1 -p err log: https://pastebin.com/9mNhyvVj

biker_rat · 10-11-2020, 09:10 AM

Two monitors at 1440p 144hz with a game running is a lot to ask from a RX 590. I am surprised that it works in Windows.

MT5496 · 10-11-2020, 10:07 AM

Quote:

Originally Posted by biker_rat

Two monitors at 1440p 144hz with a game running is a lot to ask from a RX 590. I am surprised that it works in Windows.

i tried this with a single monitor and still get the lockup, i used a nvidia GTX970 to run these 2 monitors at 1440p 144Hz and had no issue with it except immense screen tearing when running multiple graphically intense things on each monitor ( game on one YouTube on the other).

i have also tried powering the GPU with 2 power cables from my PSU but it didn't make a difference.

biker_rat · 10-11-2020, 05:57 PM

I use a RX 580 ,but I don't have your 2 monitor high refresh hi res setup. I don't know a clear path to victory for you, but I sugggest you could try:
Mesa 20.2.0?
Kernel 5.9.0 was just released 2 hrs ago.
Use adriconf to tweak mesa settings?
Try using wayland instead of X11?
Manually create an xorg.conf file?

MT5496 · 10-11-2020, 10:48 PM

Quote:

Originally Posted by biker_rat

I use a RX 580 ,but I don't have your 2 monitor high refresh hi res setup. I don't know a clear path to victory for you, but I sugggest you could try:
Mesa 20.2.0?
Kernel 5.9.0 was just released 2 hrs ago.
Use adriconf to tweak mesa settings?
Try using wayland instead of X11?
Manually create an xorg.conf file?

im using kde which has its own problems plus im relying on certain xorg features, im currently on the github version of mesa and it has improved some programs but i have figured out a method to reduce my crashed. it requires under-clocking my GPU significantly. if i under clock it then my crash problem goes away but i obviously lose performance in game. for example my GPU clocks up to 1580Mhz but will crash at anything above 1464.
im using corectrl to control the clocks. so far after 6 hours Ive had no crash in game i would like to know why this issue occurs and if it would occur with the new AMD cards because that would influence weather i stay on team red. im starting to wonder if its card specific since my XFX rx590 fatboy OC has these issues and its strictly an AMD problem for me.

i could try a new kernel but ill have to do that tomorrow.

im in the process of creating an xorg conf file but im having issues since im not familiar with its syntax.

computersavvy · 10-12-2020, 09:47 PM

Quote:

Originally Posted by MT5496

im using kde which has its own problems plus im relying on certain xorg features, im currently on the github version of mesa and it has improved some programs but i have figured out a method to reduce my crashed. it requires under-clocking my GPU significantly. if i under clock it then my crash problem goes away but i obviously lose performance in game. for example my GPU clocks up to 1580Mhz but will crash at anything above 1464.
im using corectrl to control the clocks. so far after 6 hours Ive had no crash in game i would like to know why this issue occurs and if it would occur with the new AMD cards because that would influence weather i stay on team red. im starting to wonder if its card specific since my XFX rx590 fatboy OC has these issues and its strictly an AMD problem for me.

i could try a new kernel but ill have to do that tomorrow.

im in the process of creating an xorg conf file but im having issues since im not familiar with its syntax.

That definitely sounds like you are trying to push the card beyond its physical capability. Since it works at lower speed and refresh then you seem to have a choice.
1) use it at the speed and refresh rate where it works for your current tasks. The rated speeds are for expected conditions and do not always work for extreme conditions, such as gaming.
2) Replace it with a better card that will suit your desired performance. A better card tempts us all to try and get the most out of the card, and you still might run into the "It won't do this!!" conditions.
3) Continue to try various software combinations that might improve the performance of your card, but also might not. The hassle of trying to tweak performance might be your ball game, but it definitely is not mine.

Be aware that pushing a video card beyond its design limits often results in drastically shortened lifetime. Even right at the edge can be damaging.

MT5496 · 10-12-2020, 10:25 PM

Quote:

Originally Posted by computersavvy

That definitely sounds like you are trying to push the card beyond its physical capability. Since it works at lower speed and refresh then you seem to have a choice.
1) use it at the speed and refresh rate where it works for your current tasks. The rated speeds are for expected conditions and do not always work for extreme conditions, such as gaming.
2) Replace it with a better card that will suit your desired performance. A better card tempts us all to try and get the most out of the card, and you still might run into the "It won't do this!!" conditions.
3) Continue to try various software combinations that might improve the performance of your card, but also might not. The hassle of trying to tweak performance might be your ball game, but it definitely is not mine.

Be aware that pushing a video card beyond its design limits often results in drastically shortened lifetime. Even right at the edge can be damaging.

Yeah I'm not trying to get more out of the card I'm just trying to get it to work at stock like it does in windows. In windows I don't get any crashes and it boosts to 1580 Mhz unassisted since that's it's stock boost clock. But in linux it crashes if it goes any higher than 1465. I am planing on upgrading to amds new big navi cards once linux support is irond out. So I am just hoping that I don't have this problem with that card ..
I'm just confused since I have a GTX970 NVIDIA card but it doesn't crash at all on this resolution and refresh rate even when I benchmark it.

So maybe AMDGPU driver isn't very stable with the voltages for this card? I guess I'm gonna run it underclocked until I can upgrade.

computersavvy · 10-13-2020, 06:54 PM

Quote:

Originally Posted by MT5496

Yeah I'm not trying to get more out of the card I'm just trying to get it to work at stock like it does in windows. In windows I don't get any crashes and it boosts to 1580 Mhz unassisted since that's it's stock boost clock. But in linux it crashes if it goes any higher than 1465. I am planing on upgrading to amds new big navi cards once linux support is irond out. So I am just hoping that I don't have this problem with that card ..
I'm just confused since I have a GTX970 NVIDIA card but it doesn't crash at all on this resolution and refresh rate even when I benchmark it.

So maybe AMDGPU driver isn't very stable with the voltages for this card? I guess I'm gonna run it underclocked until I can upgrade.

It is a known fact that RADEON and NVIDIA devices perform differently. Nvidia provides up to date and fully functional drivers specifically for linux, but I don't know if the same can be said for radeon devices. The manufacturer support seems different to me and AFAICT it seems nvidia is aware of the demand for video cards that work under linux so they try hard to meet the demand.
Open source drivers (nouveau for example, as compared to nvidia) seldom work as well as those from the manufacturer.

The manufacturers have a vested interest in making certain their devices just work in windows, but there seem no opportunities to check for options as we can with linux. It has been claimed that Microsoft actively discourages the manufacturers from supporting linix.

MT5496 · 10-13-2020, 07:57 PM

Quote:

Originally Posted by computersavvy

It is a known fact that RADEON and NVIDIA devices perform differently. Nvidia provides up to date and fully functional drivers specifically for linux, but I don't know if the same can be said for radeon devices. The manufacturer support seems different to me and AFAICT it seems nvidia is aware of the demand for video cards that work under linux so they try hard to meet the demand.
Open source drivers (nouveau for example, as compared to nvidia) seldom work as well as those from the manufacturer.

The manufacturers have a vested interest in making certain their devices just work in windows, but there seem no opportunities to check for options as we can with linux. It has been claimed that Microsoft actively discourages the manufacturers from supporting linix.

alright i understand hopefully i dont get this problem upon upgrading or ill end up on nvidia for linux permanently. Thanks for the help now on to fix an unrelated keyboard issue.

biker_rat · 10-14-2020, 07:15 AM

If you think the AMD driver is deficient, make bug report / ask for help from the AMD devs.
Also there are two AMD drivers, one pure opensource (except firmware) and one with proprietary elements. Although the more proprietary one is packaged by AMD and supported by AMD on a very limited number of distros, enterprising third party volunteers have repackaged it to shim it unofficially into most distros.

MT5496 · 10-14-2020, 07:40 AM

Quote:

Originally Posted by biker_rat

If you think the AMD driver is deficient, make bug report / ask for help from the AMD devs.
Also there are two AMD drivers, one pure opensource (except firmware) and one with proprietary elements. Although the more proprietary one is packaged by AMD and supported by AMD on a very limited number of distros, enterprising third party volunteers have repackaged it to shim it unofficially into most distros.

im using AMDGPU which is amds opensource driver not AMDGPU-PRO im not using the proprietary one due to incompatibility with my desktop environment. AMDGPU is built into the kernel. Maybe ill open a bug report just unsure how to do it.

obobskivich · 10-15-2020, 04:54 PM

FWIW I vaguely remember that specific model of XFX 590 card having 'known issues' relating to heat/performance (I know it has a bad reputation in the hackintosh community as well), and XFX has taken a bit of a beating in the last few generations for releasing sub-par cards in terms of reliability, thermal performance, etc (they're on revision 3 or 4 of their flagship 5700XT at this point, trying to get the cooling to work, for example). What kind of temperatures is the GPU reporting during this testing? What about the VRMs? (Is this different Windows vs linux?) It wouldn't surprise me if Windows is 'only just' not failing, and something about the different drivers/kernel/whatever is tripping your linux install up - the card is probably just not truly stable at whatever advertised OC boost clocks.

As far as 'this is influencing my future purchase' - I wouldn't touch a new launch card from either vendor for at least six months (assuming you can actually buy it at all thanks to price gougers), to let them work out whatever glitches/bugs/etc. I think the GTX 10 series (in 2016 (!)) was probably the last 'good' launch from either of these companies. Maybe Intel will do better with Xe, who knows.

MT5496 · 10-15-2020, 06:50 PM

Quote:

Originally Posted by obobskivich

FWIW I vaguely remember that specific model of XFX 590 card having 'known issues' relating to heat/performance (I know it has a bad reputation in the hackintosh community as well), and XFX has taken a bit of a beating in the last few generations for releasing sub-par cards in terms of reliability, thermal performance, etc (they're on revision 3 or 4 of their flagship 5700XT at this point, trying to get the cooling to work, for example). What kind of temperatures is the GPU reporting during this testing? What about the VRMs? (Is this different Windows vs linux?) It wouldn't surprise me if Windows is 'only just' not failing, and something about the different drivers/kernel/whatever is tripping your linux install up - the card is probably just not truly stable at whatever advertised OC boost clocks.

As far as 'this is influencing my future purchase' - I wouldn't touch a new launch card from either vendor for at least six months (assuming you can actually buy it at all thanks to price gougers), to let them work out whatever glitches/bugs/etc. I think the GTX 10 series (in 2016 (!)) was probably the last 'good' launch from either of these companies. Maybe Intel will do better with Xe, who knows.

i understand what you mean, i didnt do enough research when buying it,,, it usually gets to about 80 degrees which is high but i didnt think it was unstable high, its funny this card is my rma replacement for a different XFX 590 fatboy oc that had crashing issues in both windows and Linux but underclocking didnt help it.
and about the new gen your right i didnt plan on pre ordering it until linux support is great and there is no issues like previous gens. the 3080 crash problem definitely enforces this.

biker_rat · 10-16-2020, 09:03 AM

https://www.phoronix.com/scan.php?pa...p-Polaris-2020
Read above article. It says AMD devs have patches queued for 5.11 kernel that sound relevant to your issues.

obobskivich · 10-16-2020, 09:49 AM

Quote:

Originally Posted by MT5496

i understand what you mean, i didnt do enough research when buying it,,, it usually gets to about 80 degrees which is high but i didnt think it was unstable high, its funny this card is my rma replacement for a different XFX 590 fatboy oc that had crashing issues in both windows and Linux but underclocking didnt help it.
and about the new gen your right i didnt plan on pre ordering it until linux support is great and there is no issues like previous gens. the 3080 crash problem definitely enforces this.

I'm not absolutely certain of this, but I don't think 80* C on the top-end is 'dangerous' for that GPU (most modern GPUs can do 90-95* C peak so that doesn't sound 'wrong'), although what you've described still sounds like 'unstable OC' to me. What about the VRMs? (does it/can it monitor those? (I know not all cards can)) Look into biker_rat's suggestion, and try running it slightly underclocked in linux for the time being I guess.