GPU Price Issues

Originally Posted by you
The problem is if they run (very) hot the memory might develop latent failure. Sadly this sort of failure causes crashes and corruption under stress (parity errors); and generally not our-right failure to run.

If a GPU/VRAM is failing it usually because:

1) It was originally defective from the factory
2) Electrostatic discharge
3) Capacitor failure (not usually an issue on the better brands)
4) Someone was REALLY overzealous when OCing and went as far to use overclocking and firmware tools that went beyond the voltage safeties.

Virtually every modern GPU is designed to throttle itself with a LARGE degree of tolerance. Even if the card was under full stress, the GPU would become irrelevant for games, long before it failed. Theory and practical application are vastly different.

Originally Posted by Zloth
Yeah, I was thinking more about the lifetime of the card. If you were to get a used 1080Ti right now, you would expect it to have years left before it burns out. If it's a crypto-leftover, though, it could already have 5000 hours on it. Video cards haven't got odometers on them.

That would actually be a great feature - just store the number of compute TFLOPS or something the card has generated to date…like SSDs show the number of bytes written and have tools to monitor health. But of course they won't.

@Caddy - I work with people who run heavy GPU compute jobs 24/7. The cards die more frequently than GPUs stuck in a desktop. I would never buy a card near the end of warranty from them, because I would expect it to have issues.

Wrong

Originally Posted by Caddy
If a GPU/VRAM is failing it usually because:

1) It was originally defective from the factory
2) Electrostatic discharge
3) Capacitor failure (not usually an issue on the better brands)
4) Someone was REALLY overzealous when OCing and went as far to use overclocking and firmware tools that went beyond the voltage safeties.

Virtually every modern GPU is designed to throttle itself with a LARGE degree of tolerance. Even if the card was under full stress, the GPU would become irrelevant for games, long before it failed. Theory and practical application are vastly different.

Originally Posted by you
Wrong

No U!