Does that actually shorten its lifespan? Aren’t GPUs made to run at full power without wear? (As long as they don’t overheat, but they thermal throttle before that)
Like 87 said, thermal expansion slowly damages them, so less extreme shifts between temperatures are better. Hence when turning the PC on, I also let it slowly “warm up” before heavily loading it, and don’t run the fans too aggressively when it’s cooler.
You can also have individual components overheat and die. VRMs on some 3090s are notorious for this. I’m trying to minimize this risk by undervolting and clock capping it, so current draw never peaks too high.
I’m babying my 3090 like its the last GPU on earth. Undervolted, TDP way down, clockspeed capped, fans RUNNING.
It will never see 420W again.
Make sure you can get replacement fans before you abuse them
I can! It’s an EVGA FTW3.
Even if I couldn’t, it’s ducted to the side of an SFF case, so I could just disconnect them, wire the PWM lead from the GPU and use 120mm case fans.
The card rarely goes above 300W now and ingests ambient air, so the fans hardly spin up anyway.
EVGA, nice, all you needed to say haha
I wish they still made GPUs :(. Even an Intel Arc.
Hell, if they could somehow source 3090 dies, they’d make a killing still selling them, heh.
Does that actually shorten its lifespan? Aren’t GPUs made to run at full power without wear? (As long as they don’t overheat, but they thermal throttle before that)
That’s just textbook propaganda from hardware manufacturers.
They can say it because technically the change from cold to hot damages components… Not the actual heat… Afaik.
Also, thermal throttling tanks performance, especially in CPU’s. It causes massive hitching or even crashes.
Like 87 said, thermal expansion slowly damages them, so less extreme shifts between temperatures are better. Hence when turning the PC on, I also let it slowly “warm up” before heavily loading it, and don’t run the fans too aggressively when it’s cooler.
You can also have individual components overheat and die. VRMs on some 3090s are notorious for this. I’m trying to minimize this risk by undervolting and clock capping it, so current draw never peaks too high.