system crashes ONLY when GPU under load

[mrtexasfreedom]

I'm troubleshooting a problem I suspect is the power supply. It's a thermaltake 850 Gold 80, but don't have specific model info. Video card is R9 390X.

This system has no dust on any cooling fins and is not struggling with cooling issues. This is guaranteed.

When system is booted from cold and 3-D game is launched immediately, within moments of running, the system hard-crashes with even the power button and reset buttons not responding. I have to flip the power off on the PS, then on again before I can start up the system.

I run the Boinc client with ONLY CPU tasks, and it runs for hours at 100% CPU utilization, no crashes. I enable a GPU task and within minutes the system crashes. As long as nothing GPU-intensive runs, system is stable at idle running for days.

I do not have a second PS to test with, unfortunately. Also, no other video card.

I checked the BIOS to see what the voltages are registering as. This caught my attention:

Does that look correct? What about the other values on that listing?

Any guidance here is appreciated!

mtf

 

[Tsumi]

Motherboard readings can be unreliable. The best way is to get a multimeter and check your voltages. ATX spec for the 12v line is 11.4 to 12.6 volts. You can try watching your 12v readings with a multimeter and see what they dip to.

What is your CPU? Regardless, 850 watts should be more than enough to run any single GPU system.

 

[silk186]

I had a similar issue that I think is caused by the GPU overheating. I set a more aggressive fan profile and performance to -5%.I forget the exact setting, but it fixed my blue screen issue. It could be my PSU but I doubt that is the case.

 

[Armenius]

Motherboard readings can be unreliable. The best way is to get a multimeter and check your voltages. ATX spec for the 12v line is 11.4 to 12.6 volts. You can try watching your 12v readings with a multimeter and see what they dip to.

What is your CPU? Regardless, 850 watts should be more than enough to run any single GPU system.

I agree with this, but the sensor readings are usually not that far off of what it really is. If the motherboard is reading 10.5V then there is definite droopage going on and I'd replace the power supply. When the PSU needs to be hard reset after a shutoff event then that is the safety breaker in the unit tripping. If the PSU OP references is the one I had back in the day then it is about 15 years old, so regardless it needs to be replaced. I would never take a chance with a power supply because a $150 simple replacement could potentially turn into a $1,000+ rebuild as the failing unit takes everything else with it.