Hello
Excuse the long post, but I’ve been having trouble for about 9 months now.
My current setup is the FW13 32 gb of ram with an amd 7640u connected to a 9060xt 16gb eGPU via the adt link ut3g. I’m using a hama 80gbps cable and the psu is a seasonic 650w gold.
I’m dual booting win10 and cachyOS from the same ssd.
The issue is I’ve been having gpu crashes in both OS for a long long time now. For some reason win 10 event viewer does not show any errors, but cachyOS basically says “device lost from bus”
The crashing itself is random and seemingly sometimes affected by load, other times it isn’t. I could game for 4 hours with 0 issues, or it could crash just after boot or 30 min into a session of just the system idling. Sometimes I can work for hours and then boot up a game to get a crash after 20 min.
Logs will be posted at the bottom.
Here’s what I tried:
Switching cables and usb4 slots
Adding a fan to cool the ut3g
Aspm=off (laptop refuses to boot with the egpu connected)
all-ways-egpu (seemed to stabilise it but again the laptop refused to boot with the egpu connected)
GRUB_CMDLINE_LINUX_DEFAULT=‘nowatchdog nvme_load=YES amdgpu.runpm=0 amdgpu.gpu_recovery=1’ these are my current grub settings
Hot plugging. Sometimes hot plugging works, other times I have to boot with the ut3g powered on and connected.
Removed discord (as it seemed to crash more when I was actively using it)
FOR WIN10: if it fails to launch reinstalling the drivers from scratch seems to fix the issue for some time, I also tried using a script to disable the iGPU but that also did not work.
As a whole, win10 seems to be a bit more stable that cachyOS. Hot plugging does not work on win10.
The only 2 options I could think of and I have no investigated are:
Power supply - Im pretty sure its not overheating, the case itself gets at most slightly warm to the touch (less than body temp). I’d say enough evidence of not overheating is that sometimes I can have really long gaming sessions with no issues. The wattage is also a bit overkill.
Ut3g - same argument. The fan I’ve bought is a huge 120mm one that blows air over half of the pcb and the heatsink at most gets to body temp.
I also want to note that the card does no overheat, especially with this fan (the most I’ve recorded is 60C )
Thank you for your time, I hope someone can shed light on this.
Not as familiar with Linux. Curious if the same issues would be present under Windows 11. The AMD boards were built with Windows 11 support in mind. Not sure if that is going to make a difference here.
Possibly trying Ubuntu and Fedora on a LIVE USB to see if it is an OS thing instead of a hardware communication fault.
Try adding the following kernel parameter: pcie_ports=native pcie_ecrc=on
It will enable PCIe error reporting in Linux, and might give you a hint regarding why the PCIe disconnects.
If you do see PCIe errors, you can experiment with different cables, until you find one with fewer errors per second.
The AMD CPUs are a bit of a strange animal with its handling of PCIe errors, such that just enabling the error reporting, can actually fix the disconnect problem.
I would be interested to hear if it helps at all or not.
pkunk - Im only keeping windows because of games with kernel lvl anticheat and I’d much prefer to stay on linux as much as possible. I have considered win11 but I think it will be a last resort solution. As far as other distros, I will give it a try soon.
Morkale - Those parameters did not work for me, the problem persists.
James3 - Same thing, but I did get some new messages I have not seen:
kernel: x86/amd: Previous system reset reason [0x0800800]: an uncorrected error cause a data fabric sync flood event
I got this error when the system rebooted itself, maybe I just have not noticed it before.
kernel: acpi: pci: interrupt link LNKA configured for IRQ 0
kernel: acpi: pci: interrupt link LNKA disabled
kernel: acpi: pci: interrupt link LNKA configured for IRQ 0
kernel: acpi: pci: interrupt link LNKA disabled
repeated a few more times
Also I have not tried undervolting the egpu, maybe I’ll play around with that, but considering it crashes even when idle, I doubt it will be of help.
For the sake of trying I did turn aspm off again and the screen stays black if I boot with the gpu connected and the fans spin up and down repeatedly. If I hotplug I can still work on the laptop but the fans still go on and off. At the same time the ut3g has 3 green lights on however…
I tried forcing the gpu to a lower pci gen speed (both 4 and 3) but that did not work
Since the random crashing right now is actually instant (some days its like this, other days is more stable ) I somehow managed to have the gpu running and the ut3g is detected but I have no output on my external monitor
lact is also detecting the gpu running, so I could try lowering the clock speeds and/or reducing the wattage? I don’t know what would be acceptable though
I also found some firmware for the ut3g on the adt link website maybe there’s a more stable version ?
Please can you post the full dmesg/kernel log to pastebin or somewhere like that, and put a link to it here.
There are other messages in the log that are useful.
For example, at time 2.5 there is a Link down.
and at time 7.1 there is a Link up.
That should not be happening. It should be up all the time, and not bounce like that.
Have you tried running the laptop from battery with no power adapter plugged in apart from the eGPU?
Are there more or less eGPU problems with the laptop power adapter vs no laptop power adapter?
Sorry about the wait.
Here’s a pastebin link to a full journalctl -b -1 spanning 2 min.
What happened:
I booted with the egpu turned on, it gave some errors during the boot and then when i went to enter my password the ext monitor went on and a few seconds later it went off again, the internal screen was black, so I had to hold down the power button.
Here’s a pastebin of just the warnings and errors:
Usually when a few days pass and I try to turn the egpu on again its more stable, this time its not the case.
Interestingly enough hotplugging has the gpu running and LACT can see it, but there is no output on the ext monitor
Also to note I am on the newest framework BIOS and the newest cachyOS kernel (updated bios and kernel AFTER I made the post)
Windows 10 hasn’t been updated in about a year.
At the moment I can’t really test the gpu in a desktop PC or try to swap out the PSU.
This is a screenshot of the firmware that was on the UT3G until this point.
I tried flashing 241129 and also the newest one from april 2026, those did not work.
The most stable .bin seems to be 230525 and 230810 (cited from other threads). Of course I get an error that those cannot be flashed, so the next best thing is well, the next one - 231005, which should alsi be more stable eith amd than the new ones.
That seemed to stabilise the setup in windows and I played for about 2 hours. CachyOS still refuses to boot. I tried Fedora from a live USB, I got my hopes up when it booted and the external monitor worked, but that lasted for about 5 minutes until the whole system rebooted itself…
I’ve tried unplugging the HDMI from the egpu to run the connection back to the internal screen, that did not work. Display port makes no difference as well.
For the heck of it I also tried amdgpu.reset_method=4, still nada.
I again managed to play for 2 more hours with 0 interruptions.
This firmware I flashed seems to keep the drivers stable, because I now have sound coming out of my monitor.
This is something I failed to mention before, but usually when I get my setup to work, the next boot the sound just wont come out of the monitor itself, I remember seeing some related errors in cachy but I did not mind them.
Hey, sorry to bother you, but someone on the egpu.io forum a while back told me I could tag you if I keep having issues.
If you could shed any light on this that would be awesome. Thanks!
I remembered I have a working “ancient” 1050TI bought during the high scalper prices 4-5 years ago.
Installed some drivers and plugged it in (I’ve actually used it before but on windows to test the adt link when I got it), got an image on my external monitor, it crashed and the fans spun down. Fans spun up and I got a connection again. This happened 1 or 2 more times and It stopped working.
Here’s a pastebin of the errors…
окт 05 20:18:46 cachyos kernel: amdgpu: Overdrive is enabled, please disable it before reporting any bugs unrelated to overdrive.
Don’t turn on overdrive.
окт 05 20:19:01 Framework13 kernel: pcieport 0000:62:00.0: Unable to change power state from D3cold to D0, device inaccessible
окт 05 20:19:01 Framework13 kernel: pcieport 0000:63:00.0: Unable to change power state from D3cold to D0, device inaccessible
…
окт 05 20:19:12 Framework13 kernel: NVRM: GPU at PCI:0000:64:00: GPU-a9c71b45-c572-8c93-bec8-6ff97a09cf36
окт 05 20:19:12 Framework13 kernel: NVRM: Xid (PCI:0000:64:00): 79, pid=2907, name=kwin_wayland, GPU has fallen off the bus.
окт 05 20:19:12 Framework13 kernel: NVRM: GPU 0000:64:00.0: GPU has fallen off the bus.
окт 05 20:19:12 Framework13 kernel: NVRM: Xid (PCI:0000:64:00): 154, GPU recovery action changed from 0x0 (None) to 0x1 (GPU Reset Required)
I would contact NVIDIA about this to find out what is going on with the device falling off the bus. This could be something wrong in their driver stack for eGPU usage.
I turned on overdrive a while back to see if I can squeeze a bit more performance out of the setup but when it crashed I just reverted back the settings and all was well, so I gave no second thought about the setting itself, especially since the crashes were happening way before I enabled overdrive.
Turned it off and now everything works.
Stressed it for about 1 hour and it seems stable.
Thank you, I did not expect something so simple to be the cause, but I will continue testing for now.
Crashed after 2 hours, had it both under load and then just watching videos.
It managed to reconnect once or twice for a couple of seconds but then the device was dropped and the laptop did not freeze this time.
Here are the error messages:
These are the error messages when I tried to boot again:
Whats interesting is that HDMI crashed during boot but display port works, I’m currently typing this with display port connected.
7.2.9-1-cachyos
Hardware config? The 9060xt. The NVDIA card I just plugged in to rule out the gpu itself.
Used a local LLM in the morning for some work, was fine during the 15-20 min I needed it, then crashed while idling 40 min later, I think the AC was plugged in the same side however (front left port AC, the back left is where I always plug the USB4 cable of the UT3G).
Later the same day I managed to game for almsot 4 hrs, no crashes, but the AC was on the other side now.
Maybe that affects stability as well, haven’t really noticed it before, even though usually the AC is plugged on the back right.
Will continue testing in a few days as I will be working.
)
) I somehow managed to have the gpu running and the ut3g is detected but I have no output on my external monitor
