The core clock is undervolted while mining - but the memory clock is absolutely cranked to the maximum it will support.
I've been idly mining on an RTX3080 for 6 months now, and the hashrate has slowly gone down about 20% since new. I suspect the thermal pads on the memory chips are slowly cooking. I replaced them once already and restored a bit of hashrate.
? My CPU actually underclocks itself if I am running a heavy AVX-512 workload. Actually, back when I had it in a cheaper motherboard, my system would crash because the MB couldn't deliver enough power to the CPU.
But I agree with the first sentence. Lots of miner cards are undervolted because what matters is performance per watt, just like many server cpus.
It underclocks, but generates more heat in the AVX-512 case (that’s the whole reason for the underclocking). If miners are undervolting to reduce power consumption, heat generation is going to lower by necessity.
It's not a myth. These things get racked in stacks of GPUs in what may or may not be actually-cooled data centers. Undervolted or not, when you couple that with constant load you're stressing devices manufactured to consumer specs.
This is mostly a myth, miners cards are usually undervolted, and only a subset of the card is used.
If your CPU only used it's floating point unit then it wouldn't be as intense as if you were using the whole thing.
The only potential issue is thermal damage, but that's hard to place and has to be close to, or in excess of 90-110c.