Screen freezes with GPU artifacts when RX 580 GPU is under heavy load

ilikeeatingtacos97 · 2022-08-12 17:44:13

Whenever I put my GPU under heavy load, such as playing a video game, my screen will freeze with GPU artifacts.
The computer keeps working, I hear audio of videos continuing to play, but all screen output is completely frozen.
Typically it's a random completely solid color, or random pixels, or it will be what was last on the screen but completely distorted.
Most of the time it's a random solid color.

The GPU works fine so long as it's not under heavy load. And it is not overheating when the screen freezes.

I ask here because I think it may be a drivers issue, rather than a hardware one as I've had an extremely similar problem with another GPU by AMD (R9 280) on this same computer and Arch install, meanwhile an Nvidia GPU I have (GTX 560) has no problem.

My graphics card is an RX 580 8GB and I am running the AMDGPU drivers. Kernel version is 5.18.16-arch1-1

Journalctl error log from just as it freezes, errors seem like some storage problem but it only occurs as I get this screen freeze.
https://pastebin.com/42XEmcue

ilikeeatingtacos97 · 2022-08-12 21:49:11

Since posting this, I had another freeze while testing. Once again while playing a video game. But this time with a much more detailed log.
This time, not only did audio continue playing in my browser, but I could see my mouse cursor and it was undistorted. I was even able to click to pause and unpause the video.

The visuals even continued to update a couple times after it froze, still horribly distorted, but being sorta able to make out what was on screen. It stayed frozen eventually, but with my mouse cursor still continuing to move on screen completely undistorted.

Log file from this freeze:
https://pastebin.com/zjjGF4E8

Lone_Wolf · 2022-08-13 17:34:28

Looks like a low level issue.

What brand & model is your system/motherboard ?
Which bios firmware version ?

Are you using X or wayland ?

Please post full outputs of $ lspci -knn and a journal of current boot .

Note : pastebin.com is not very popular here,
See https://wiki.archlinux.org/title/List_o … n_services for alternatives .

Last edited by Lone_Wolf (2022-08-13 17:35:41)

ilikeeatingtacos97 · 2022-08-14 00:27:30

Motherboard is a GA-F2A88XM-D3H
BIOS version: F8
BIOS Revision: 4.6

I am using X.

Oh, sorry.
Output from lspci -kkn:
https://pastebin.osuosl.org/view/e3185b27

By journal of current boot, you just mean my current boot? I haven't had the issue in the current boot but here it is:
https://pastebin.osuosl.org/view/80fd0538

I noticed the issue doesn't seem to occur if I undervolt my GPU and limit it's MHz to like 600.

Lone_Wolf · 2022-08-14 11:04:48

Often there are symptoms of low level issues visible in dmesg/journal before the issue happens.

gigabyte lists 2 hw revisions for that board, 3.0 and 3.1, both appear to use the same bios[1] .

F9 and F10a are later versions then your installed one, and there's

Aug 13 17:24:21 mycomputer kernel: mtrr: your CPUs had inconsistent fixed MTRR settings
Aug 13 17:24:21 mycomputer kernel: mtrr: your CPUs had inconsistent variable MTRR settings
Aug 13 17:24:21 mycomputer kernel: mtrr: probably your BIOS does not setup all CPUs.
Aug 13 17:24:21 mycomputer kernel: mtrr: corrected configuration.

Although I doubt this has to do with the gpu issue, it does indicate firmeware issues.
For now keep the F8 version.

A very minor issue :

Aug 13 17:24:21 mycomputer kernel: Command line: BOOT_IMAGE=/vmlinuz-linux root=UUID=face77ed-87f0-4a6f-b66d-48922ab3acd3 rw cryptdevice=/dev/sda2:cr_root radeon.si_support=0 amdgpu.si_support=1 loglevel=3 radeon.si_support=0 amdgpu.si_support=1 mitigations=off

The RTX 580 doesn't need the radeon.si_support or amdgpu.si_support options as it will always be driven by the amdgpu kernel module .
It shouldn't do any harm though.

On to the graphic artifacts issue .

Your description and undervolting + limit gpu frequency not causing them makes this smell like a power issue.

That can be caused by hardware ( PSU lacking amperage, faulty cable, bad connector etc) but also by software .

Aug 13 17:24:23 mycomputer cpupower-gui[599]: Applying configuration...
Aug 13 17:24:23 mycomputer cpupower-gui[599]: Setting CPU: 0
Aug 13 17:24:23 mycomputer cpupower-gui[599]:     Minimum Frequency: 1700.0 MHz, Maximum Frequency: 3700.0 MHz
Aug 13 17:24:23 mycomputer cpupower-gui[599]:     Governor: Powersave, Online: True
Aug 13 17:24:23 mycomputer cpupower-gui[599]: Setting CPU: 1
Aug 13 17:24:23 mycomputer cpupower-gui[599]:     Minimum Frequency: 1700.0 MHz, Maximum Frequency: 3700.0 MHz
Aug 13 17:24:23 mycomputer cpupower-gui[599]:     Governor: Powersave, Online: True
Aug 13 17:24:23 mycomputer cpupower-gui[599]: Setting CPU: 2
Aug 13 17:24:23 mycomputer cpupower-gui[599]:     Minimum Frequency: 1700.0 MHz, Maximum Frequency: 3700.0 MHz
Aug 13 17:24:23 mycomputer cpupower-gui[599]:     Governor: Powersave, Online: True
Aug 13 17:24:23 mycomputer cpupower-gui[599]: Setting CPU: 3
Aug 13 17:24:23 mycomputer cpupower-gui[599]:     Minimum Frequency: 1700.0 MHz, Maximum Frequency: 3700.0 MHz
Aug 13 17:24:23 mycomputer cpupower-gui[599]:     Governor: Powersave, Online: True

Do the artifacts also occur when you use the default schedutil governor ?

[1] https://www.gigabyte.com/Motherboard/GA … rt-dl-bios

ilikeeatingtacos97 · 2022-08-14 23:15:29

I've used this PC for a long time, never had issue with my GPU until a little while ago.

As in, my CPU schedutil governor?
Yes. Just tested it, still happens.

Generally I have it at powersave when I boot PC, and increase it to ondemand or performance when doing something demanding.

seth · 2022-08-15 07:00:13

I've used this PC for a long time, never had issue with my GPU until a little while ago.

100% of working stuff is not broken before it is broken…

Try the LTS kernel and also to pass

pcie_aspm=off amdgpu.dpm=0 amdgpu.aspm=0 amdgpu.runpm=0 amdgpu.bapm=0

to the kernel, https://wiki.archlinux.org/title/Kernel_parameters

ilikeeatingtacos97 · 2022-08-20 22:57:21

I've tried using those kernel parameters, and I do not get the graphical artifacts if I use them.
However GPU performance takes a very heavy hit with any demanding application.

I've tried them with both the ordinary Kernel and LTS Kernel. Both don't have the issue if I use those kernel parameters.

Log when using LTS Kernel with those parameters:
https://pastebin.osuosl.org/view/ec89d823

seth · 2022-08-21 06:09:25

Is there a parallel windows installation? Do you overclock the GPU (or anything else)?
Those parameter should™ prevent the GPU from powering down, so "GPU performance takes a very heavy hit" seems unreasonable.

Check the actual performance levels, https://wiki.archlinux.org/title/AMDGPU#CLI_(default) w/ and w/o the GPU and you can also try https://wiki.archlinux.org/title/AMDGPU … cy_problem instead.

ilikeeatingtacos97 · 2022-08-21 23:24:13

I do not have a paralell Windows installation, and I do not overclock anything.

I meant just that I get worse performance in any game I run with those parameters without any undervolt, compared to running without parameters and undervolting my GPU.

With the kernel parameters, none of the files for checking GPU utilization work except for mem_info_vram_used and mem_info_vram_total.
If I attempt to read them with cat I get "No such file or directory"

But when I remove those kernel parameters, all of them work just fine except for /sys/class/drm/card0/device/pp_od_clk_voltage.

Still I tried to check GPU utilization with a game open both with and without the parameters.
I used a program called Radeontop that seemed to get at least some data about my GPU with the parameters active. And I used both Radeontop and those commandline commands for the GPU utilization data without the parameters.

And the non-parameter data was with my GPU being undervolted to 29 watts.

GPU utilization data:
https://pastebin.osuosl.org/view/713b84a5

Also my monitor isn't above 60Hz, it's exactly 60.

giostark · 2022-08-22 07:36:22

As I can read around (you are not the only one with this issue) seems PSU or kernel. Or faulty VGA.
45 to 55 are idle temps. Are you sure that during the load the temps stay good enough ?
Software side I can't say a thing , but maybe is worthy considering the dust. (because the problem occurred after long time)
Is the VGA well cleaned? If you have a compressor or a spray can of air compressed try to blow in all the space under the heat sink (laterally). (if you use the spray can , remove the power!!! - it condensate - )
The dust moved by the fans is very subtle and if in a good amount can conduct a little electricity (because contain minerals). Ah...clean also the PSU.
I'm curios to see how will be solved this.

ilikeeatingtacos97 · 2022-08-25 23:11:19

Well I replaced my PSU [XFX TS 550W 80+ Bronze] with a new one [EVGA 600 BR, 80+ Bronze 600W].
Same issue as before.

I am very thorough with my dusting, I have an electric duster just for it. I gave my PC another round of dusting while I was replacing my PSU, and I'm still getting the issue.

My idle GPU temp is, I think, like 38. It got up to like 47 before it froze last time I attempted it.
Seems for some reason my GPU's fans aren't actually turning on by default, I have to turn them on with a program called CoreCtrl.
But even with them on, and my GPU not reaching a high temperature, I still get it.

I'm doubtful it's my GPU as of the three I have tried, two are having the exact same issue.
All three were plugged into the same slot on my motherboard so I don't think it's that either.

No idea what else to try.

giostark · 2022-08-26 09:22:53

I try to be cooperative...
So , just for understand the situation.
Two AMD VGA have this issue. One Nvidia don't. Right?
Two faulty power supply is too rare to be an option.
Seems compatibility between mobo , VGA , linux .
I know could be bothering to try the follow. Trial and error...
I would try:
1) To be sure of this and have no more doubt about the functionality of the two VGA, you could try a windows copy on a usb stick and run some bench or game. (detaching the linux drives for avoid that winzoz write on them during the installation)
The AMD drivers on win should be fine so you would be sure that is a linux software side problem.

I'm a noob software side so please reconsider what I'll write if could be a possible option.
2) Could you run the applications that freeze just from X ? (avoiding the desktop environment) ("sudo systemctl disable sddm" , reboot , startx , and then run the application?)
Could be an Xorg problem? (or anything in between the kernel and the application?)

Lone_Wolf · 2022-08-26 09:30:07

Have you tried removing the indivdual kernel parameters one at a time to verify which combo solves the issue ?

(there's a good chance the parameter(s) causing the performance drop is not needed to solve the artifacts)

Last edited by Lone_Wolf (2022-08-26 10:02:48)

seth · 2022-08-26 11:50:57

Also

Also my monitor isn't above 60Hz, it's exactly 60.

is irrelevant to the approach - the idea is to keep the GPU in a performance mode (low, high will likely immediately cause the freeze?) instead of disabling PM altogether.

The performance drop might be caused by the GPU defaulting to it's minimium settings.
With a max draw of 185W and no issue when undervolted, I'd however rather say that the PSU is insufficient (or you forgot the 6/8-pin dedicated power connector, but that's less likely - maybe it requires two adapters and you provided only one?)
The GTX 560 draws a max of 150W (~80%)

=> Try to limit the GPU to 800MHz (stable?) and 900MHz(on the edge?) and 100MHz (probably too much?)

ilikeeatingtacos97 · 2022-08-31 17:10:03

Lone_Wolf wrote:

Have you tried removing the indivdual kernel parameters one at a time to verify which combo solves the issue ?
(there's a good chance the parameter(s) causing the performance drop is not needed to solve the artifacts)

I tried all the paramters on their own, all either having the performance drop or giving artifacts.

I don't know how to quote more then one post but,
"The GTX 560 draws a max of 150W (~80%)"
I used to run my R9 280 with just a 550 watt PSU just fine. As far as I can tell, the RX 580 uses less power then the R9 280. And now I'm using a 600 watt PSU.
I don't know how accurate it is, but a program I've been using called CoreCtrl is detecting my RX 580 using at moe 50 watts before I get artifacts.

seth · 2022-08-31 19:49:22

It's not only about the PSU, the power also has to get there.

seth wrote:

=> Try to limit the GPU to 800MHz (stable?) and 900MHz(on the edge?) and 100MHz (probably too much?)

You can insert random quotes, https://bbs.archlinux.org/help.php#bbcode and copy and paste the text.

Arch Linux

#1 2022-08-12 17:44:13

Screen freezes with GPU artifacts when RX 580 GPU is under heavy load

#2 2022-08-12 21:49:11

Re: Screen freezes with GPU artifacts when RX 580 GPU is under heavy load

#3 2022-08-13 17:34:28

Re: Screen freezes with GPU artifacts when RX 580 GPU is under heavy load

#4 2022-08-14 00:27:30

Re: Screen freezes with GPU artifacts when RX 580 GPU is under heavy load

#5 2022-08-14 11:04:48

Re: Screen freezes with GPU artifacts when RX 580 GPU is under heavy load

#6 2022-08-14 23:15:29

Re: Screen freezes with GPU artifacts when RX 580 GPU is under heavy load

#7 2022-08-15 07:00:13

Re: Screen freezes with GPU artifacts when RX 580 GPU is under heavy load

#8 2022-08-20 22:57:21

Re: Screen freezes with GPU artifacts when RX 580 GPU is under heavy load

#9 2022-08-21 06:09:25

Re: Screen freezes with GPU artifacts when RX 580 GPU is under heavy load

#10 2022-08-21 23:24:13

Re: Screen freezes with GPU artifacts when RX 580 GPU is under heavy load

#11 2022-08-22 07:36:22

Re: Screen freezes with GPU artifacts when RX 580 GPU is under heavy load

#12 2022-08-25 23:11:19

Re: Screen freezes with GPU artifacts when RX 580 GPU is under heavy load

#13 2022-08-26 09:22:53

Re: Screen freezes with GPU artifacts when RX 580 GPU is under heavy load

#14 2022-08-26 09:30:07

Re: Screen freezes with GPU artifacts when RX 580 GPU is under heavy load

#15 2022-08-26 11:50:57

Re: Screen freezes with GPU artifacts when RX 580 GPU is under heavy load

#16 2022-08-31 17:10:03

Re: Screen freezes with GPU artifacts when RX 580 GPU is under heavy load

#17 2022-08-31 19:49:22

Re: Screen freezes with GPU artifacts when RX 580 GPU is under heavy load

Board footer