Ryzen freeze or AMDGPU+GCN 1.1 bug?

[dc740]

Small update for people coming to this post from google . This was finally fixed, as you can see here:

https://bugs.freedesktop.org/show_bug.cgi?id=91880

The fix consist in <step 1 from previos post>, and upgrading the kernel. Good luck!

[dc740]

same problem with gcn 1.0 + ryzen 1700 + ubuntu 18.04 generic kernel 4.15.4

problem first appeared when i i upgrad to ryzen 1700

All I can say is:
1) Update your motherboard BIOS, and search for a setting like this: "Typical Current Idle". Make sure you change it to ON (do not leave it in "Auto". It must be manually enabled)
I have a Gigabyte motherboard, and the setting can be found here:
M.I.T->
Advanced Frequency Settings ->
Core Settings ->
*Power Supply Iddle Control = Typical

The names can be different from one motherboard brand to the other.

2) Edit /etc/modprobe.d/blacklist.conf and add "blacklist amdgpu" to the bottom. After that run "sudo update-initramfs -u -k all". This way you will make sure it won't be loaded.


I was able to ahve a 100% reliable system by doing those two steps. The first step "fixes" this very public and widespread bug: https://bugs.launchpad.net/ubuntu/+s...x/+bug/1690085
The second step (I think) is related to the power management with AMDGPU on some cards (I own an AMD 390x, which is GCN 1.1). I'm going to enable AMDGPU again right after they release kernel 4.19, which is supposed to have new power management features (and I hope they fixed this bug as a side effect)

[paupav]

same problem with gcn 1.0 + ryzen 1700 + ubuntu 18.04 generic kernel 4.15.4

problem first appeared when i i upgrad to ryzen 1700

[dc740]

The Xfce power manager and screensavers can cause freeze, i noticed that several months ago. The xorg server has dpms built in, my screen blanks in 30 minutes or so without the xfce power manager and a screen saver installed. My system wakes ups nicely. You can control dpms with xorg.conf, see Arch Linux wiki.

I'm using LXDE, but it has exactly the same problem and had to do the same workaround before.

About this issue: I used 4.17rc3 for a few weeks, and I no longer saw the hangs while going idle and turning off the screens. However, I got random lockups and no panic messages. Even REISUB didn't work. So I went back to my reliable fix: blacklist AMDGPU and use radeon only. It's very disappointing, but it's the only thing that seems to work. The power management does work OK with the radeon module. From what I remember when I started with this thread, the fix only works while AMDGPU is completely blacklisted and it hangs if it's loaded, even if it's not in use.

[skeevy420]

I don't know if this will help, but I've experienced a similar system lockup with my R7 260x and AMDGPU and not with Radeon (processor is Dual Xeon X5687). What was causing my lockups was the "Lock the screen when going to sleep" tick box in the XFCE Power Manager. Unticked that and my desktop is working great again. That's not a good solution if your PC is somewhere public. I'm the only person who uses my PC or who has access to it so it's just fine for me. I'm not sure if it's a Light Locker, AMDGPU, or XFCE bug, but I know how to trigger it -- let it idle, blank, security kicks in, and then wait more than 20 mins...if it just kicks in and I catch it, it doesn't freeze up, but if it sits for more than 10 or 20 mins, it'll freeze up.

What sucks is I'm having such great results with gaming on AMDGPU these days that I'd rather deal with this bug versus switch drivers.

[pete910]

Duel bios does not work in gigabyte b350 gaming 3. I got my money back from that shit.

So, No then

[pete910]

Buy Asus mobos only. I did return the gigabyte b350 gaming 3 mobo after updating the "fail safe" bios and the mobo did not boot anymore. Asus Prime B350 series is good.

Don't know how to use the duel bios then take it

Built several without issue

[dc740]

Quick update:
I upgraded to Ubuntu 18.04LTS and also installed kernel 4.17rc3. Then I white-listed AMDGPU and configured grub to use it again. The system is rock solid. I left it several times alone to go idle and it didn't freeze. It also successfully shut down both monitors when it had to (It used to hang while trying to do this).

Thanks!

UPDATE: check posts below. It freezed again, and I had to upgrade the BIOS to get a new setting and fix it there.

[dc740]

Hi folks

I'm planning to buy an AMD 2700x to be used as Continuous Integration Server with Debian Testing(basically to run tests of a big app written in Ruby), so it won't have a desktop environment. Do you think this bug could affect me also? Should I disable C6?

Thanks!

Yes, it may affect you. I haven't recompiled the kernel yet, but the problems seems to be quite random. The bug reports show some people that still have instability even when disabling C6. As a developer myself, I have to admit it's hard to tell whether it's related to Ryzen or something else on every user case.

Many people will report their success with one method, and that method may not be the same for other users with different motherboard, or different hardware. Adding the kernel parameter worked for debianxfce, but there are cases where the kernel parameters didn't help (I have to assume they didn't know they had to recompile the kernel to enable them).

For me, I "fixed" it by disabling AMDGPU, and now the system is rock solid. I've been using it intensively for the last 5 days without any problem, and I used to have at least one freeze a day. For future reference, I have the latest BIOS installed (F22), and IOMMU=soft in the kernel parameters (because the BIOS implementation is buggy and it didn't like my sound card using the hardware IOMMU).

[007lva]

Hi folks

I'm planning to buy an AMD 2700x to be used as Continuous Integration Server with Debian Testing(basically to run tests of a big app written in Ruby), so it won't have a desktop environment. Do you think this bug could affect me also? Should I disable C6?

Thanks!