How reliable are modern CPUs?

ochinko writes: Big IT companies report that they see a growing number of irreproducible CPU errors that are due not to architectural design flows, but to the limits to which semiconductor manufacturing was pushed.

Computer chips have advanced to the point that they're no longer reliable: they've become "mercurial," as Google puts it, and may not perform their calculations in a predictable manner.

...

Lately, however, two of the world's larger CPU stressors, Google and Facebook, have been detecting CPU misbehavior more frequently, enough that they're now urging technology companies to work together to better understand how to spot these errors and remediate them.

Error correcting code?

[Z00L00K]

Seems to me that we will need not only memories with ECC, but also CPUs and probably also control circuits.

Some of the errors might be due to quantum effects that starts to show when we now get so small and narrow, some from pure thermoelectric noise and some from outside radiation. And I doubt that we can work around that in any other way than having more bits to build up an error correcting design. Of course we can also add liquid nitrogen cooling, but that comes with new problems.

Re:

[lobro1]

In addition to needing error correcting code, another possibility might be badly written software which raises the operating temperature of of CPU unnecessarily . I've noticed this on my home desktop as I monitor the temps of the CPU cores. Microsoft routine update caused this when Edge was installed (crammed down my throat). Edge was difficult to remove (by design?) Thanks MS