[Solved] [HELP NEEDED] eGPU setup that was working stopped working completely.
I need help in troubleshooting the issue.
macOS 11.6.1 via OCLP 0.3.3 with Kryptonite and disabled dGPU.
I unplugged the cables and plugged them back in. This cause the whole setup to fail. Before this the external monitor would cut out but after a reboot it worked again. So it seemed something was about to fail and unplugging the cables was the final straw.
Suspected the cables were the culprit so connected another mac on and copied data back and forth. Data copies fine. TB2 cable seems ok.
Had no spare TB3 -> TB2 adapter so bought another. Still no go.
In system report the Akitio Node shows up in Thunderbolt/USB4 but the eGPU does not. How to I go about troubleshooting from here?
Resitted the GPU in the eGPU box and everything's working again. The eGPU box has never been moved or touched. It solved it though.
@psypryss, Yah, I get that sometimes. It happens because the machine puts some sort of request to the device and if I had to guess it gets stuck on the Thunderbolt controller. When I look up the event in system view with the Thunderbolt lanes it says "error 0x00000bc4" which is a failed operation instruction. A hard reset of the unit normally resolves it.
@psypryss, Normally I just unplug mine for 10 seconds. I hate removing cards from PCI-E sockets. I always feel like i'm going to break them.
@psypryss, I have a similar setup (same MBPro + Node) tho I'm running 10.14 for legacy apps. Have you made any changes to your AKiTiO Node? I'm running the same fans and firmware since I bought it new in 2017.
@cnon297, No. The eGPU just stopped working after I unplugged the cables from the eGPU box. The node detects but the GPU doesn’t. Removing it from the PCI slot. Cleaning the contacts a bit made it work.
I had the same issue: working eGPU setup with Nvidia and hot-plugging that started failing every time.
I was able to get it working again by removing the video card from the PCIe socket in the eGPU box, and disconnecting the video card's molex power connector, and then reconnecting it all again, like the OP had done.
Before that, I tried unplugging the power cable to the eGPU box's power suppy, then re-plugging, but that didn't seem to fix it. Although, I might not have waited for enough time after booting to see if the card was plugged in the 20 times I tried it.
I was hot-plugging properly each time, plugging the eGPU's active Thunderbolt 3 cable to the left side of my 2017 MacBook Pro w/ Touch Bar fully loaded immediately after the circles start spinning under the Windows logo during boot. If you don't get the cable fully plugged in within half a second, it can fail.
After Windows logs in and you want to check if the eGPU is working, it's good to rely on more than just running "Nvidia Inspector", because it can remain blank for a while after Windows fully logs in. Running "Nvidia Control Panel" will have a definitive error message with a red "X" icon if it's not working, or will show a different error with the word "you" in it if it's going to work but just needs more time, and then a few seconds later trying to run it again should show the control panel. (I will update this comment with screenshots of the errors when I can reproduce it.)
- Operating system: Windows 10 Pro 10.0.19042 Build 19042
Chocolatey installed software versions (reported from `choco list --local-only`):
- nvidia-display-driver 497.29
- nvidia-inspector 184.108.40.20600706
- nvidia-profile-inspector 220.127.116.11
@nhooey, Hmm, that just seems odd. If the unit is completely discharged then I don't see how removing the card from the break out board would make a difference. Shrug... If the power supply has a flip switch, I would try that first, count to ten, and restart the laptop, then try again. See if that works.