They are all pre autotune units running original firmware at the original clock rate. My fail miners (one r4 and one s9) hosting the failed boards were updated to the autotune software for giggles and testing purposes but that was after the boards failed running on their original hardware and software.

They are in a cool room with air temps under 80F year round, and a clean, dry environment with stable power. Let's just say I have a bunch of other equipment (About half a dozen servers and GPU setups) that have had a 0% failure rate in the same room on the same power.

Different equipment same conditions, one with an obviously higher failure rate.

Of course comparing the QC of a company like IBM or HP to Bitmain is kind of apples to oranges. One obviously doesn't put the same care into their product to test long term quality. I'd prefer they raise prices a little and raise the QC bar, but it's their business and their choice.

I'm sure if I get enough quality servers and GPUs I'll run into a failure at some point, it can't stay 0% forever. But those companies also offer much better warranties and servicing if that happens.