I'll be trying to reinstall the drivers with a complete wipe shortly. If this makes no difference, I'll definitely consider RMA. My card is still under warranty, but is this kind of "non-crippling" issue (I can work around it but it is very annoying and sort of dangerous) eligible for an RMA?
I'm still not sure how the card itself could be bad, considering it's failing at low temperatures and working well at higher temperatures. This particular thermal characteristic pretty much rules out any kind of software issue, but honestly I don't see what hardware could be causing this issue - it can either be the PSU, the motherboard, or the card itself - any other component has no direct hardware communication with the GPU. Nevertheless the CPU is fine, and my RAM's passed memtest86+ 100 passes (both sticks, checked that a few weeks ago).
The PSU's been eliminated from the equation, I'm really not sure how a failing motherboard would cause this confusing thermal reaction from the card, so it's really only the GPU left. Or could it be the motherboard?