Hi guys, I'm facing a continuous problem with my current setup:
I've changed my old Sapphire R7 240 4GB video card with a newer XFX RX580. Since then I have had a lot of problems whenever I use any proglam that works with GPU: e.g. Adobe PS/LR, VLC viewing videos, even a browser like Chrome or Firefox when there is some page with embedded videos. My PC starts to show random errors regarding memory access, disk corruption and so on, until ilt crashes with a BSOD (normally with CRITICAL_PROCESS_DIED error).
Also, all diagnostic/benchmarking programs like (but not only) GPU caps viewer does the same disaster... I can replicate the situation all times
In any case, before crash I'm no longer to launch any program, nor ctl-alt-del works, neither shudown completes.
What i did:
- swapped video card with another one (identical) - no good
- completely reinstalled Windows 10 from scratch - no good
- removed AMD drivers in safe mode: the problem arises also without AMD drivers, just in case... - still no good
Other than swap the card with another one that does not use AMD drivers, there is anything else I can do? Any ideas will be appreciate a lot.
Thanks in advance!
Seems that your psu is defective, try with another one. If the system still crashes maybe it's a memory related problem, then you should test your memory modules with memtest. But i would try swapping the power supply first.
Windows has a built-in memory diagnostic tool. Start with using that.
At this point, it could be memory, CPU, or PSU issue. Just need to test each in isolation to narrow it down.
Memory was thoroughly tested, with M$ diagnostic and other programs, never return some kind of error... The PSU seems more than sufficient for all the ensemble but I'll give a try.
I have connected the only connector I can find on the card... unfortunately there is no manual explaining anything!
I forgot to mention that the same tests running under Linux OS has worked for many many hours without any hassle...
So the PSU should not be the problem.
Anyway, I've connected the 8-pin connector straight to an 8-pin connector of my PSU (it should carry ~ 150w) instead of 2 x 6-pin connector with the Y-shaped cable. AFAIK, 6-pin connector are capable of carrying ~ 75w of power. The result should be the same, right?
Tonight I can try another PSU, identical to the one I'm using at the moment... I can exclude memory modules problems, I've tried them one by one, no difference.
BTW, using 2x6-pin instead of 1x8-pin connectors gave same results: crash on first stress test...
Have you set your power limit to plus 50? RX 580's have their power ceiling set too low by default as the vanilla drivers are set to reference specs and almost all the cards sold have some factory OC to them. I have the XFX card too. Raise the power limit and set a custom fan curve. It should help your stability. If not then I would suspect a hardware issue. GPU, Motherboard or Power Supply. Is your bios on the mother board the latest? Here are my RX 580 settings, use the power limit and fan curve and see if that helps.
- tried a new PSU (identical to the previous one), no difference
- tried the newest driver, 19.5.1, no difference.
- tried the above settings (thanks pokester!) , no difference
Memory and PSU seems to work ok on another PC, and in fact we should remember that all stress test made under Linux OS (Mint 19) has worked flawless for many hours and hours
I'm out of ideas at the moment, so I'm getting rid of the RX 580. What card should I buy? Nvidia? Other AMD?
Try one last test in Mint to discard a memory fault. Install google stressapptest and run it 1 or 2 hours.
- sudo apt-get install stressapptest
and to run 1 hour:
stressapptest -W -s 3600 -M
to run 2 hours:
stressapptest -W -s 7200 -M
If it passes try with another card, i'd try with sapphire.
And one last suggestion, maybe it's a problem with your graphics card memory. Try to downclock the gpu memory to test if the system is more resilient to errors.