4 ms·

Major chip design errors, of this magnitude, are exceedingly rare. Especially in comparison to the software industry where bugs are released to production a cou

by whack 9y ago

Major chip design errors, of this magnitude, are exceedingly rare. Especially in comparison to the software industry where bugs are released to production a couple orders of magnitude more frequently. How many other Intel hardware bugs can you point to that are comparable to the ones below? "All computers with Intel chips from the past 10 years appear to be affected... The security updates may slow down older machinery by as much as 30 percent, according to the report." "discovered a problem with how early versions of Intel’s Pentium chip calculated numbers... Intel had to recall some chips and took a charge of more than $400 million."

emn13 9y ago
That's what you're saying (and indeed what I said too!), but that's not what bloomberg published. What you're saying makes sense! But they didn't qualify this by major. And the qualification is important to the heart of the story, since the way they phrased it suggests that chip design is largely reliable and thus that this is an exceptional error, when in fact errors are common, and impact is merely exceptional. In effect: they're overstating the signal-to-noise ratio this event implies. Worse, they understated the noisiness of the chip error "signal" but they also understate the noisiness of the stock value signal. And hey presto; if you conveniently present only the data you want all of the sudden a correlation looks really meaningful! Furthermore, even the qualification "major" isn't really deserved quite to the extent that bloomberg implies; specifically there have been worse chip errors much more recently than 20 years ago (e.g. amd's somewhat similar phenom issues). What makes this one worse isn't just the bug, it's how chips are used: cloud computing makes this much more relevant than similar bugs not too many years ago.
- someguydave 9y ago
  Now you know how to weight the truth of mass media publications.
gsnedders 9y ago
> How many other Intel hardware bugs can you point to that are comparable to the ones below? Disabling a whole feature (TSX) across an entire generation of processor and early steppings of the next? That's pretty major, given it's a marketed feature.

by whack 9y ago

emn13 9y ago

That's what you're saying (and indeed what I said too!), but that's not what bloomberg published. What you're saying makes sense! But they didn't qualify this by major. And the qualification is important to the heart of the story, since the way they phrased it suggests that chip design is largely reliable and thus that this is an exceptional error, when in fact errors are common, and impact is merely exceptional. In effect: they're overstating the signal-to-noise ratio this event implies. Worse, they understated the noisiness of the chip error "signal" but they also understate the noisiness of the stock value signal. And hey presto; if you conveniently present only the data you want all of the sudden a correlation looks really meaningful! Furthermore, even the qualification "major" isn't really deserved quite to the extent that bloomberg implies; specifically there have been worse chip errors much more recently than 20 years ago (e.g. amd's somewhat similar phenom issues). What makes this one worse isn't just the bug, it's how chips are used: cloud computing makes this much more relevant than similar bugs not too many years ago.

someguydave 9y ago
Now you know how to weight the truth of mass media publications.

gsnedders 9y ago

> How many other Intel hardware bugs can you point to that are comparable to the ones below? Disabling a whole feature (TSX) across an entire generation of processor and early steppings of the next? That's pretty major, given it's a marketed feature.