[Bug 219399] System panics after several hours of 14-threads-compilation orgies using poudriere on AMD Ryzen...

[bugzilla-noreply at freebsd.org]

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=219399

Don Lewis <truckman at FreeBSD.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |truckman at FreeBSD.org

--- Comment #56 from Don Lewis <truckman at FreeBSD.org> ---
I'm seeing similar problems here.
  Ryzen 1700X
  Gigabyte AX370-Gaming 5
  64GB Crucial DDR-2400 ECC RAM (from Gigabyte QVL)
  Seasonic 650W Prime Titanium PSU
  NVIDIA GT21x video card (VESA text mode only for the console, no X11,
  and no nvidia driver).

I originally started with a Gigabyte AB350-Gaming motherboard but swapped it
out due to high VRM temps and lack of ECC RAM support.  Both boards ran four
passes of memtest86 and overnight prime95 runs w/o error.  Thee first board had
the
original BIOS that didn't know about the 20C Tctl offset, so it always ran its
fans very fast.  The F5 BIOS that I installed on the X370 board had the AGESA
1004a update and knew about the Tctl offset.  I've been seeing idle temps in
the BIOS of about 35C on it.

Both boards either randomly lock up (blank console) or silently reboot (no
panic messages in /var/log/messages and no crash dumps), along with some other
errors while building ports with poudriere.  I was getting a bunch of the
"unable to rename temporary" errors while running an older version of FreeBSD
12.0 (r320316?), but haven't seen that since upgrading to r320570.

Things I've tried since upgrading to the new motherboard:
  Set fan speed at max
  Core performance boost -> off
  SMT -> off
  RAM 2400MHZ-> 1866 MHz
  CPU 3.4GHz -> 3.0GHz
  Disable 4 cores (auto -> 2+2) - Seems to run 2x longer at half throughput
    before rebooting (which happened very early in the morning when the
    room temp was at its lowest).  Exhaust air was very cool.
  Dragonfly patch

I just upgraded to the latest (F6) BIOS which has AGESA 1006.  I should have
some results in a few hours, but at this point, I'm not optimistic.

In the meantime, my AMD FX-8320E is rock solid under the same sort of load, but
 just takes too long to build all the packages sets that I use.

-- 
You are receiving this mail because:
You are the assignee for the bug.