Web lists-archives.com

Re: [WARNING] Intel Skylake/Kaby Lake processors: broken hyper-threading




Am Montag, den 26.06.2017, 08:34 +0000 schrieb Holger Levsen:
> On Sun, Jun 25, 2017 at 09:19:36AM -0300, Henrique de Moraes Holschuh
> wrote:
> [...]
> > Apparently, Intel had indeed found the issue, *documented it* (see
> > below) and *fixed it*.  There was no direct feedback to the OCaml
> > people, so they only found about it later.
> 
> [...]
> > We do not have enough information at this time to know how much
> > software
> > out there will trigger this specific defect.
> > 
> > One important point is that the code pattern that triggered the
> > issue in
> > OCaml was present on gcc-generated code.  There were extra
> > constraints
> > being placed on gcc by OCaml, which would explain why gcc
> > apparently
> > rarely generates this pattern.
> > 
> > The reported effects of the processor defect were: compiler and
> > application crashes, incorrect program behavior, including
> > incorrect
> > program output.
> > 
> > 
> > What we know about the microcode updates issued by Intel related to
> > these specific errata:
> > 
> > Fixes for processors with signatures[1] 0x406E3 and 0x506E3 are
> > available in the Intel public Linux microcode release
> > 20170511.  This
> > will fix only Skylake processors with model 78 stepping 3, and
> > model 94
> > stepping 3.  The fixed microcode for these two processor models
> > reports
> > revision 0xb9/0xba, or higher.
> > 
> > Apparently, these errata were fixed by microcode updates issued in
> > early
> > April/2017.  Based on this date range, microcode revision 0x5d/0x5e
> > (and
> > higher) for Kaby Lake processors with signatures 0x806e9 and
> > 0x906e9
> > *might* fix the issue.  We do not have confirmation about which
> > microcode revision fixes Kaby Lake at this time.
> 
> so in conclusion: don't buy intel. At least in future.

In conclusion: don't buy AMD or Intel CPUs.

AMD processors have bugs too. Just look at the recent changes to the
amd64-microcode package. If you are unlucky, there might not even be a
workaround and it can take months to reproduce, find the root cause and
fix it. Comparing the patch IDs can give you a hint how many iterations
were needed.

I don't see much difference between AMD and Intel when comparing
Henrique's announcement with my first-hand experience with one of the
vendors.

> I must say I'm utterly disappointed by this crap. "hey there is a hug
> bug, we
> dont tell you what it is exactly, or how we fixed it, but YOU MUST
> INSTALL THIS
> BINARY BLOB TO FIX IT. (and btw, this is for skylake, for kaby lake,
> ahem, maybe,
> we have no idea, but do install that crap^wblob too.")

The same complaint can be said about the AMD microcode updates.

-- 
Benjamin Drung
System Developer
Debian & Ubuntu Developer

ProfitBricks GmbH
Greifswalder Str. 207
D - 10405 Berlin

Email: benjamin.drung@xxxxxxxxxxxxxxxx
Web: https://www.profitbricks.com

Sitz der Gesellschaft: Berlin.
Registergericht: Amtsgericht Charlottenburg, HRB 125506B.
Geschäftsführer: Achim Weiss.

Attachment: signature.asc
Description: This is a digitally signed message part