News for hackers. Stuff that matters.

A comment on Ryzen CPU HyperThreads break if 100% busy and interrupted to top of Memory. Parent comment

Re:Ryzen CPU HyperThreads break if 100% busy and interrupted... by gruez on Friday August 04, 2017 @10:58PM (#14933990)
> Their response is going to be "if it crashes your application, turn SMT off".

it happens when SMT is disabled

>Epyc is on a new stepping of the silicon, it's possible this is already fixed on it. (Threadripper is not, however)

that's assuming they caught this bug, which i doubt is the case because it's only discovered now rather than being documented in the errata.

[ Parent | on HN ]
Re:Ryzen CPU HyperThreads break if 100% busy and interrupted... by dom0 on Saturday August 05, 2017 @06:46AM (#14935133)
I dunno, there were several issues in the SMT implementation publicized earlier, it is entirely possible that the root cause is the same or related.
[ Parent | on HN ]
Re:Ryzen CPU HyperThreads break if 100% busy and interrupted... by shizzlest on Saturday August 05, 2017 @07:51AM (#14935297)
its possible they overlooked mentioning it in the errata but still fixed it with their stepping. dont give up last hope :)
[ Parent | on HN ]