Comment 197 for bug 1690085

Revision history for this message
In , b-o-s-s (b-o-s-s-linux-kernel-bugs) wrote :

As a test, I've been running like below with an uptime of 12 days without any freeze or strange syslog entries. Note that this is with the patch Seth mentioned far above _reverted_, so at least for me this seems to have fixed the issue. I've read that for some it hadn't, strange. I have to countercheck another few weeks.

# uname -a
Linux donald 4.13.12-x64 #14 SMP Sun Nov 12 17:23:57 CET 2017 x86_64 AMD Ryzen 7 1700 Eight-Core Processor AuthenticAMD GNU/Linux
# zcat /proc/config.gz |grep RCU_NO
  CONFIG_RCU_NOCB_CPU is not set
# cat /proc/cmdline
BOOT_IMAGE=/boot/vmlinuz-4.13.12 root=/dev/nvme0n1p3 ro
# ./zenstates.py -l
P0 - Enabled - FID = 8C - DID = 8 - VID = 3A - Ratio = 35.00 - vCore = 1.18750
P1 - Enabled - FID = 87 - DID = A - VID = 50 - Ratio = 27.00 - vCore = 1.05000
P2 - Enabled - FID = 7C - DID = 10 - VID = 6C - Ratio = 15.50 - vCore = 0.87500
P3 - Disabled
P4 - Disabled
P5 - Disabled
P6 - Disabled
P7 - Disabled
C6 State - Package - Enabled
C6 State - Core - Enabled

Some relevant dmidecode output:

Base Board Information
        Manufacturer: ASRock
        Product Name: AB350 Gaming-ITX/ac

BIOS Information
        Version: P3.10
        Release Date: 08/28/2017

Memory Device (x2)
        Data Width: 64 bits
        Size: 8192 MB
        Type: DDR4
        Type Detail: Synchronous Unbuffered (Unregistered)
        Speed: 3200 MT/s
        Part Number: G.Skill F4-3200C14-8GFX
        Rank: 1
        Configured Clock Speed: 1600 MT/s

I'm running RAM on its specified XMP profile 3200-14-14-14-34 1.35V, which is quite good actually. UEFI settings are mostly default, only P-state overclocking is used and two memory related options turned off: GearDown mode and Bank Swapping. Voltages are default except CPU offset -100 mV. All turbo and power saving stuff is working as intented using ondemand governor.

So... are you guys sure your freezes are no UEFI setting and/or kernel config and/or hardware, most probably RAM, issue? Especially RAM is really critical on this platform and maybe Memtest cannot detect it...?
(But as noted, I have to countercheck at least 2 weeks.)