I bought the AMD Radeon RX 5600 XT PCIe card for my new desktop PC I recently built, and while it does work, there have been some pretty annoying issues with it.

The Problems:

If I go to something like Google Maps, the display will very likely freeze up completely to the point where I can't even change to another virtual terminal (ie Ctrl+Alt+F2), as the screen is completely frozen. It's virtually guaranteed to happen if I zoom in to fast, in fact every time I zoom in while looking at a map the display freezes up completely.

Another issue is that video playback when using something like VLC can be very choppy. Also if I connect to my wifi access point while playing a video it can (and often does) cause choppy video playback while my machine is connecting to my wifi access point.

There is also the issue of HDMI audio constantly cutting out at regular intervals, if I watch a video and route the audio through HDMI rather than using the onboard audio hardware. The audio does come back after it cuts out, but misses whatever sound there was after it's cut out and cut back in again.

What I've found out through some research I've done about it:

Through some investigations I've done, it seems I'm far from alone and AMD's amdgpu has some issues with the newer Navi based cards, including their Windows driver, not just AMD's Linux driver. So I'm pretty sure it's bugs in the amdgpu driver that are causing my problems. I did find this that describes my problem exactly and mentions the hardware I have (well the 5700 XT, but the 5600 XT I have is pretty much the same, just with less graphics RAM).

My system specs:

Video card: Sapphire Radeon RX 5600 XT Pulse 6GB
distro: OpenMandriva Lx 4.1
kernel: 5.5.12-desktop-1omv4001

Some things I've tried:

I have tried setting the PCIe version in the BIOS to PCIe 3.0 from the "auto" setting, since the card supports PCIe 4.0, but that didn't change anything. I also tried updating the system BIOS, but that didn't change anything either.

According to the link I posted above regarding the Google Maps/screen freezing problem at least that issue is apparently fixed in kernel 5.6, but no update to the kernel package has been released for OpenMandriva last time I checked. I don't really want to compile the kernel myself and would prefer to use the stock package instead. I have been thinking about changing distro's (and still am) but it's a PITA to have to do that, and the time to do it considering the post-installation stuff I do is an issue as well.

I must say it's pretty disappointing to have these issues with an AMD video card under Linux, since the main reason I bought it was because AMD's drivers for Linux are open-source.

Does anyone know if these issues still exist in kernel 5.6, or have any thoughts about this?

Have you checked the graphics card firmware? This article suggests you may need the latest:

https://www.phoronix.com/scan.php?pa...ux-vbios&num=1

Also I see two sites recommending the 5600 XT for Linux, so some configurations are working well. Sorry yours isn't working. What motherboard is in your desktop?

No, I haven't tried updating the firmware on the video card itself. There was an update for the kernel's firmware package though, but I still have the same problems last time I checked. I think I've seen the link you posted above when I was researching my issues just after I bought the card, but I have no idea how to update the firmware on the card itself though.

My motherboard is the ASUS Prime Z390-P.

You may not need to update the card if it has loadable firmware in the driver.

The problem with motherboards that I've run into is that they may share lanes across multiple slots so only certain combinations of cards work properly. I'm failing to understand your mother board spec, but I'm guessing the card is OK in the 1st x16 slot. Not sure about the 2nd.

The other possibility is that ASUS has done something to "work better" with their own ASUS-branded ROG_STRIX card (i.e. locking out competitors), although I've never heard of them doing that.

https://rog.asus.com/graphics-cards/...-gaming-model/

Welcome back jsb,

You're not going to like this, but I'm going to suggest that is exactly what you need to do - first. You can't fight with any kind of driver bug if you haven't installed the latest stable kernel (at the time of writing 5.9.13). Your 5.5.x kernel has been EoL'd so that's a complete dead end.

In some cases, with very new hardware it can be a good thing to also build then mainline kernel, just to see if the issue has been resolved upstream, but that's another matter. I suggest you compile and boot into the latest stable kernel first to see if the problem goes away.

1 members found this post helpful.

Are you talking about PCIe lanes or HSIO lanes? As while I'm no expert, my understanding is that HSIO lanes connect devices to the chipset and the PCIe x16 graphics port's lanes connect directly to the processor, but HSIO lanes can be shared between more than one device, and only one device can use an HSIO lane at a given time (ie. not more than one device can be using the same HSIO lane at the same time).

The board has, from memory, a grey PCIe port which is the x16 speed port (for graphics cards) that my video card is connected to, with the other PCIe port for graphics unused. I did have another PCIe card installed in the second PCIe port for graphics (the long PCIe port's) and did move it to another PCIe port to see if it would solve the problems, but it made no difference at all.

The board also supports AMD's "CrossFire" technology for having two video cards installed - AMD's answer to NVIDIA's SLI technology (I don't intend on making use of it though).

I suppose it's possible, but I'd also think it's probably unlikely.

Thanks cynwulf.

Yeah, I'm pretty much resigned to the fact that I'm not going to see the end of my problems with kernel 5.5 I'm currently using. I was thinking about seeing if I find find a live image of a distro that already has at least kernel 5.6, and then seeing if that might give any clue as to whether my problems have been fixed or not. But it would have to be both a distro that has at least kernel 5.6 (preferably a higher version), and has at least some multimedia apps and codecs installed in the live image also.

At this point, and beyond your suggestion of compiling the kernel and my idea of trying a live image of something suitable, it looks like the only other option is to put up with it until my distro actually releases a better kernel or I change distros instead.

It's PITA either way it seems...

Well, I figured I may as well try and compile the latest stable kernel (kernel 5.9.14) from kernel.org, but it looks like it's failed to build right at the end of the build process.

This is the error I get:

I did lookup the error, but I'm not sure what I'm supposed to do to fix it. I did find the following links, but while they look like patches, I don't know what to do from this point on.

https://github.com/OpenMandrivaAssoc...kernel-rc.spec
https://lore.kernel.org/linux-kbuild...t2@tip-bot2/T/

I tried again to compile the kernel, but got the same error. So I've just tried the "mainline" 5.10-rc7 kernel and I'm still getting exactly the same error as above. The rest of the kernels at kernel.org are not suitable to even bother trying to compile.

Does anyone have clue as to what the hell it's talking about, and exactly WHAT I'm supposed to do to fix it?

I'm wondering if you could rpm install a Mageia kernel? v8 beta has 5.9.12 or newer. v7 has 5.7.19 in updates.

Well it seems that your distribution uses LLVM/clang as the base compiler. You may need to seek advice from distribution's official forums / mailing lists.

1 members found this post helpful.

It's really tempting to try some rpm packages from another distro, it really is...

I did try setting the CONFIG_UNWINDER_ORC kernel option to CONFIG_UNWINDER_ORC=n in the .config file, but it kept getting changed back to CONFIG_UNWINDER_ORC=y every time I run the make command to build the kernel. But I couldn't figure out which kernel option was responsible for doing it. So I'm going to try one more thing and see what happens. So it's compiling again for at least the forth time now, so we'll see if it actually builds vmlinuz (the kernel itself) this time.

I'll let ya's know what happens.

Nope, still exactly the same error.

But I did find this that seems to suggest that LLVM Clang 9.0 can compile the mainline kernel without any patches. But ironically enough it says that at the time of writing amdgpu had some issues, yet it did in fact seem to at least compile the object code for amdgpu every time I tried to build the kernel.

So I'll see I can find anything about building the kernel from source on OpenMandriva, and failing that I guess I'll have to contact OpenMandriva or create an account on their forum to ask on there (or ask on the kernel's forum/mailing list). As I'm buggered if I know what else to try now.

As binary kernels from other distributions are most likely not built with clang, but gcc I wouldnt bother with those. Definitely try mainline next.

1 members found this post helpful.

Yeah, I'd prefer to either compile the kernel myself, or get an updated version through the package manager. I did also try building the mainline kernel as well, but got exactly the same error.

In any case, I setup an account on OpenMandriva's forums and posted a thread about compiling the kernel from kernel.org to that. So hopefully I can get some answers from there, and let ya's know if and when I can get the bastard built. The only issue will be that, since in that case I've built it myself, I won't have a kernel-devel package to install from the package manager for it. So I'll need to check what headers are in the kernel-devel package for my current kernel and find the same files in the source for the kernel from kernel.org, or figure out how to install the needed header files for the kernel from kernel.org so I can re-compile the drivers for Virtualbox for the new kernel.

I'll let ya's know.