Just turn off speculative execution.
My desktop is 8 years old. In the last couple weeks, the graphics driver has been crashing. I fear the old GPU may be nearing the end.
I paid $400 for my graphics card 8 years ago. Similar-benchmarking cards cost⌠about $400 now. Thanks a lot, coins.
Why would the driver crashing make you think the hardware is broken? Itâs much more likely to be so many other things. Try booting a clean OS with the newest driver and see if that has issues.
I canât reliably reproduce it, it seems random as far as I can tell. The whole machine freezes hard, for 30+ seconds at a time. Mouse does nothing. Ctrl-alt-del and ctrl-shift-esc both do nothing.
To get an error message I have to go to Event Viewer. Googling the error seems to mostly give people not getting answers from 5+ years ago.
I have updated the driver, and rolled back to a couple old versions.
Updated windows as well. If by clean OS you mean safe mode or a fresh install, I havenât tried that.
You donât even need to try fresh install. You can boot Linux off a USB stick. If itâs a hardware issue, your computer should crash no matter what software itâs running.
It could also be a disk issue, corrupted files and such.
I would look to a RAM hardware issue before a GPU issue if its causing a systems crash. GPU hardware failure is much more likely to cause the wildest artifacts you will ever see on screen for a bit before it forces your system to shutdown from overheating. I had this happen to my graphics card years after I owned it and at the time (5 years ago) it was under a lifetime warranty with XFX.
Memtest doesnât find any problems.
Running an Ubuntu live USB, it crashes running glmark2
, 4 out of 5 times so far. Similar behavior: hard freeze, syslog reports graphics driver errors.
I suppose I can take the GPU out and run the test again.
I am wondering, though. The PSU is also 8 years old. What are the odds the GPU is ok, and the power supply is failing? That might be hard to test, I have a bunch of SSDs in this machine and not much else to load the system.
Ok, now it is sounding more like there is definitely a hardware problem.
It could be the GPU itself. The way to confirm/deny that is to try out the video card in a different computer and see how that goes.
If the PSU is the problem, then itâs also not impossible that the PSU has caused damage to other parts of the computer. It is indeed hard to test. Youâll have to check the voltage and current on each pin of the PSU outputs and such.
Itâs also possible that itâs a GPU-adjacent problem. Are you using a DisplayPort cable? Could be the pin 20 problem maybe? Could be PCI-X slot is somehow flaky? Could be incorrect configuration in UEFI/BIOS? Lots of possibilities.
Even if the GPU fails in a different computer, you also want to confirm that it isnât something about your computer that caused the GPU to fry because it will just fry the new one you put in there as well.
Piggy backing on Scottâs post, I second running the PC videocardless if your mobo has a VGA or Displayport or something using your CPU. Follow up with separately testing the GPU in another machine as Scott said to see if it has issues.
The GPU has 1 HDMI and 1 DisplayPort, I was using them both.
I unplugged the DisplayPort, stress test has now completed without crashing 8 out of 8 times. I guess that narrows it down.
I only have the one DisplayPort cable. Is monoprice trustworthy to not fry anything?
https://www.monoprice.com/product?p_id=37920
Update: it was the cable. Why does DisplayPort feel like a return to 2002 technology? Itâs supposed to be the good one.
Weâve discussed this before. DisplayPort is really good. Thereâs just many cable manufacturers out there that have wired up cables incorrectly.
That wasnât even my problem. Iâm pretty sure this was the one that came with the ultrasharp. It was fine for like 4 years.
Sure, cables fail. But a cable causing the entire system to hang feels very not-robust in an era of âplug it in and it just worksâ.
That kind of thing can happen with any cable, especially a cable that carries power as opposed to just data. There are plenty of cases where bad USB stuff messes up computers. Youâre actually lucky it wasnât the pin 20 problem specifically as it would be likely to permanently damage the computer.