Hacker News Reader

I believe the concern was that it ran on GPUs (mining on CPUs hasn't been cost-effective for quite some time). Since GPUs are rarely run constantly and don't see loads like mining produces during typical use, their owners might have overclocked them under those assumptions. With those assumptions violated, I wouldn't be surprised to see cards fail prematurely.

View on HN

Don't even need to overclock them. Given a sample size of 14000 GPUs, I would expect a couple hundred to at least bluescreen simply from being placed under a sustained serious load for the first time.

Try running the FurMark stress test on any given machine. If you don't get glitches or a reboot, that GPU is above-average stability in my experience.

This is a sad comment on the state of consumer hardware. Speaking from my viewpoint as a computational chemist, a video card that fails memtestG80 is not merchantable and gets slung back at the manufacturer.

After a while you have learned how fast companies deal with RMAs and you know that factory-overcooked hardware is next-to-useless.

>This is a sad comment on the state of consumer hardware

And also incorrect. A system crashing on a combined Linpack/Furmark load is defective or contains a defective component. Even extremely thermally-limited laptops do not suffer from this due to protection mechanisms.

Reply on news.ycombinator.com