Feature #11975
closedUpdate AMD CPUID for F17 M10,30,70
100%
Description
We should update the list of AMD socket detection logic and chip revision logic for AMD family 17h models 10h-1fh, 30-3fh, and 70-7fh. This covers Ravens Ridge (some Zen+), Matisse (Zen 2 client parts), and Rome (Zen 2 Server parts).
Updated by Rick V over 3 years ago
I've found your zen branch and patched it against the current illumos-gate
and the patches in #11953, will report back with any changes.
Updated by Robert Mustacchi over 3 years ago
Hey Rick,
I was waiting until we finished dealing with all the B450 issues before dealing with this. I cleaned up a bunch of what was in that branch, so there's still a bit to do there. I should get the actual versions up for review (there were some minor things I had to clean up from what was there).
Updated by Robert Mustacchi over 3 years ago
Review for this is at https://code.illumos.org/c/illumos-gate/+/277.
Updated by Robert Mustacchi over 3 years ago
On a Zen+ platform we now see the correct socket (tested by Gary Mills):
$ psrinfo -vp The physical processor has 4 virtual processors (0-3) x86 (AuthenticAMD 800F11 family 23 model 1 step 1 clock 3100 MHz) AMD Ryzen 3 1200 Quad-Core Processor [ Socket: AM4 ]
Updated by Robert Mustacchi about 3 years ago
I went through and tested this on an AMD Rome system by manually printing the cpuid information. The reason that this doesn't show up in psrinfo -vp is because I have hyper threading and for some reason it doesn't.
Updated by Electric Monk about 3 years ago
- Status changed from New to Closed
- % Done changed from 90 to 100
git commit c0692105a8b350050833b89a3c492068ed009b09
commit c0692105a8b350050833b89a3c492068ed009b09 Author: Robert Mustacchi <rm@fingolfin.org> Date: 2020-04-03T05:16:40.000Z 11975 Update AMD CPUID for F17 M10,30,70 Reviewed by: Richard Lowe <richlowe@richlowe.net> Reviewed by: Jason Lippert <jason@glasspelican.ca> Reviewed by: Toomas Soome <tsoome@me.com> Approved by: Joshua M. Clulow <josh@sysmgr.org>