eGPU crashes repeatedly after working flawlessly for a few months - XPS 15 9560 + Aorus 1080 Gaming Box
My current setup:
Aorus 1080 Gaming Box
I've had this setup for more than a year now, for the first couple of months it was working perfectly fine. However, for the past 3 months, I have been getting crashes while using various applications (from 3d work to some light gaming). Most of the time, the eGPU would just restart, closing any applications that were using the GPU, after awhile the eGPU reconnects and everything is back to normal (until it decides to do it all over again). Sometimes the whole laptop would goes black, unresponsive and I would have to restart, and occasionally, the eGPU would even refuse to reconnect after restart and I'd have to discharge the box for it to work. This would happen almost daily, oftentimes multiple times per day.
Some more info:
I'd probably have to note that the problem starts to appear around the time I removed my battery, and started running the laptop on AC primarily. Also, it seems to happen whenever I plug both my Dell adapter and the USB-C into the machine (which should be fine according to my own research, however odd since it wasn't like this before). When I run only the USB-C cable to my laptop, it runs fine but the performance is capped since the cable only delivers 65W, when the laptop needs 130W to perform properly.
I tried monitoring my system before the crashes, everything seemed normal (GPU temps ~50oC, CPU temps ~78oC). I don't really know if it's related but it would crash more often if I turn on Turbo Boost on my CPU.
I tried installing NVIDIA drivers, both the latest versions and the older ones, but to no avail.
There was once I tried to clean the outlet by blowing out the dust, the problem seized for 3 days but then it came back again :(.
I'd greatly appreciate any effort given to help troubleshoot my problem, I tried to just ignore it but it has gotten more and more on my nerves the past couple weeks. I tried so many things and I think I'm losing my mind every time it happens. I apologize if I miss any info and would be happy to provide any more related to this problem!
Thank you all!
@ddwscmin, Do you have the Nvidia dGPU enabled while the eGPU is connected? I would say to start with running DDU to remove the existing drivers first. Next is to disable the dGPU then install drivers with eGPU connected.
@itsage, Update: I've done everything you told me to, I tried going for one night but it didn't seem to fix anything, I got the crash after 10 minutes of gaming (Turbo Boost not even enabled, CPU and GPU temps were way below 75oC and 45oC). 🙁
@ffoggy, I've tried to look for the registry but I can't find it? Maybe the settings is not on my laptop?
Anyways, is it likely that the problem could've been caused by the USB-C cable or the PSU? I've been looking at other forums and some say that those two are the culprits. Is there a way that I can troubleshoot such problems, since supply is pretty limited where I live and the Thunderbolt cables as well as the flex PSU can cost a lot.
Edit: The cable that I'm using is included with the LG Ultrafine display one, the specs seem pretty nice but IDK if it's possibly the cause to all this.
@ddwscmin, Try adding a TdrDelay ValueData of 10 and a TdrDdiDelay ValueData of 20 in the Windows Registry Editor, under: HKEY_LOCAL_MACHINE\System\CurrentControlSet\Control\GraphicsDrivers
You can also try changing NVIDIA Global Power Settings to Prefer Maximum Power.
For further information on TDR, I suggest reading this article.
@ddwscmin, Perfect! Let me know if that works.
@boltoway, Update: I've been running flawlessly for just about a day since I tried your method, but sadly it just decided to crash, unfortunately 🙁 Before that the laptop even completely shuts off while I was rendering in Blender, causing the eGPU to not connect afterwards. After having everything up and running again it crashed after I started the rendering again (with Turbo Boost on, I was turbo boosting previously after the "fix" but it seemed to ran fine, little did I know).