(I remember when Intel got a lot of shit because their floating point module in early Pentiums had some bad cells in a lookup table for doing division. And that was an error that only showed up in certain instances, and even when it did, it was usually correct to the first 6 to 8 digits IIRC.)

Yeah this is the example I always use too

This event spawned one of greatest niche jokes ever internetted: “I am Pentium of Borg. Division is futile, you will be approximated”

(One of my engineering professors would rant about this, on occasion, as something that was both stupid to happen (because it was frankly a dumb error in the PLD) and way overblown (because in most instances the error would be way less than measurement error anyway.).

As a kid I loved the joke that ran in MacWorld (or maybe MacUser) magazine: “Sure it cost $475 million to fix the problem, but who’s counting? … not Intel”

And I’ve just realized/remembered that, given the time, I was told that joke by an actual, physically present human (vs reading it on social). God I’m old 😐

(Ok, it was going to bother me so I had to go look up the lookup error. Worst case errors were up to the 4th sig. digit. That was about one case in a billion. Intel fixed it (eventually) because a computer that can't do math correctly is a bad joke.) en.wikipedia.org/wiki/Pentium...

And Intel profited in the medium term from being seen as a company that stood behind its products. Now, companies that tell their customers to enjoy their shit sandwich only go up in valuation.

(The failsons hadn't completely conquered every branch of industry in the US in the late 90s. Nowadays, tho... ugh.)

Ah yeah I read about that in this blog - super cool analysis www.righto.com/2024/07/pent...