I had some time to test some more. Just tried starting the game but not actually going beyond the main menu. My computer crashes in the same time interval which is a bit odd. I’m not sure if this signifies that the real issue is not based on true workload. I also tried using the amd software under tuning just to see if under volt or overclock helped ( they did not and actually made the situation worse so I am trying to bring everything back)
I’m assuming from the image of your setup that the GPU cable is connected to the thunderbolt 4 cable. Do you run it without the power monitor normally (the power monitor was just for testing)?
If you run it with the power monitor, it is possible that is interfering.
One thing you could try, although I don’t know how much it would help, would be to just take out the USB C expansion card and plug it directly into the motherboard.
This would only help if the issue was caused by signaling though. It is also possible that your Thunderbolt cable is damaged but you might want to try this first.
@ Elliot_Lu, Yes to the Thunderbolt 4 cable. I did not purchase the power monitor until the suggestion in this thread to check power.
I did some testing and did not find any appreciable difference with or without the power monitor or with or without the usb c expansion card. I also tried a few different thunderbolt cable and saw no difference.
what did make a drastic difference is connecting my monitors to my PC through a usb hub (ankerA8385) (Amazon.com: Anker USB C Hub, PowerExpand 11-in-1 Docking Station, with 4K@60Hz HDMI and DP, 100W Power Delivery, USB-C and 3 USB-A Data Ports, 1 Gbps Ethernet, 3.5mm Audio, microSD and SD Card Reader : Electronics) as opposed to through the GPU itself but have the GPU connected with no monitor.
Since I was seeing performance on the UT3G similar to the K43SG+ ssd enclosure I switched back to the K43SG+ssd enclosure set up (with the monitors connected to the hub rather than the back of the GPU as they were when this thread started). This way I can get my money back on the UT3G and am fairly happy with the graphics performance.
it does still bother me that I don’t fully understand what is was causing the crashes though.
When you connect it to through the dock, it is likely either using the iGPU or routing through the iGPU this is probably why you do not experience the crashes. You will lose some performance, but it is better than crashing. Just making sure, have you looked for any newer driver for eGPU you are using?
This might not be related the your GPU setup at all. What RAM are you using?
I’m facing issues with my Framework 13 AMD laptop crashing too, the screen goes black and then reboots.
I’m using Team Elite 5200MHz RAM (2x 32GB), and apparently non-5600MHz RAM can be a source for instability when the system is connected to power. I don’t have the issue running on battery, which unfortunately you’re unable to test. Just letting you know as might not be due to your GPU enclosures, etc.
@Elliot_Lu I am using the AMD Adrenalin software to keep up to date. It checked this morning for updates and did not find any. The current software is 24.9.1 released on 9/26/24
@Andrew_B I purchased the ddr5 5600 ram directly from framework (2x 8GB cards). but you description for crashing matches what I was seeing
Hey! Not sure if I should start a new thread because I have similar issues.
While playing games (and ONLY when playing games) either screen freezes with graphical glitches:
or just hard reboots the pc.
Happened 4 times in the last hour. I have eventviewer logs if anyone is interested, I can’t get much out of it (skill issue).
Not using a eGPU though. Just normal FW13 7840u with 32gb RAM (Crucial 5600).
Was checking the temps of the pc, because it was blowing air like mad, but they never really exceeded 85ish degrees.
Also tried stresstesting with 3Dmark with no issues. Also tried with or without using the (FW original) powerbrick.
On my wits end with this
I’ve since acquired a stick of Crucial 32GB 5600MHz RAM and still got the crashing.
I also had it happen on battery this morning. I was in a meeting, when I came back my computer had rebooted. So for me the issues happens regardless of whether power is connected or not, but much less likely to occur when running on battery.
Sorry you are having issues narrowing down what the cause is. Some thoughts came to mind as I was reading this over the last few days.
- It could be a thermal issue with the CPU. Could be a poor paste application, or it is not making even contact so at high temps over a long period of time the EC on the board panics and shutsdown.
- The power supply could be “negotiating” at the wrong time and the power supplied suddenly drops, creating artifacts and panic from the board. In a battery powered laptop this would be mitigated from the battery acting as the buffer.
- The power powering the laptop (I know great english here) is suppose to be fed from the dedicated power supply, if the laptop thinks it can get power from another source (the dock?, the eGPU?) and it switches (remember only 1 USBC port can be a source for Framework mainboards) then realizes there actually is not enough power and the voltage level drops suddenly the machine will panic and shutdown.
Those are just some of the thoughts going through my head as I read this. One way to crudely test the thermal issue of the CPU is taking a hairdryer (set it to COLD/NO Heat) and put that baby up against the fan in the Cooler Master case. If it goes a decent amount longer before resetting itself that would help narrow it down to a thermal issue with the paste, heatsink, or the amount of even pressure making contact with the CPU. (there have been a few of those known to crop up in the forums)
@pkunk those are interesting ideas. the hairdryer test saw a little improvement in nova bench but not in actual game (it acted the same as if now hairdryer was there). Items 2&3 seem possible, I wonder if there is a way to tell which USBC port the mainboard is pulling power from or if there is a way I could log that value and see if it switches just before shutdown. I also wonder why the issue almost completely goes away if I connect my monitors to my USBC hub rather than the graphics card itself (even though the graphic card is clearly doing the graphics work based on both the Adrenalin software and the game quality itself)
The issue likely disappears because the dGPU output is being routed through the iGPU. When you plug anything in through the laptop’s I/O rather than directly from the dGPU, it will always be routed through the iGPU in the case of the Framework 13. This can lead to reduced performance, but maybe in your case, the reduced power draw or heat might be preventing it from crashing.