4P G34 Machine Check Reboot

[plext0r]

Hello all,
I'm still tweaking my "new" GL motherboard with 4 spicy 6282 chips (natively 2.3GHz). I've had them clocked at 3.4GHz on 1.1875V for a few weeks pulling around 620K PPD on 8101 WU. I figured that since I've now got the VRMs sinked, I could try going further.

This is on air inside a Blackhawk Ultra case.

I bumped to 3.5GHZ on 1.1875V and it seemed stable for a few hours, so I bumped to 3.6GHz on 1.2V. I received the following error and the machine rebooted about 15 minutes later.

Mar 19 22:04:39 gkar kernel: [Hardware Error]: CPU:21 MC1_STATUS[-|CE|MiscV|-|AddrV|-|-]: 0x8c00000000110151
Mar 19 22:04:39 gkar kernel: [Hardware Error]:  MC1_ADDR: 0x0000000000000001
Mar 19 22:04:39 gkar kernel: [Hardware Error]: Instruction Cache Error: Decoder uop queue parity error.
Mar 19 22:04:39 gkar kernel: [Hardware Error]: cache level: L1, tx: INSN, mem-tx: IRD
Mar 19 22:04:39 gkar kernel: [Hardware Error]: CPU:21 MC5_STATUS[-|CE|MiscV|-|-|-|-]: 0x98000000000c0e0f
Mar 19 22:04:39 gkar kernel: [Hardware Error]: Execution Unit Error: DE error occurred.
Mar 19 22:04:39 gkar kernel: [Hardware Error]: cache level: L3/GEN, mem/io: GEN, mem-tx: GEN, part-proc: GEN (no timeout)
Mar 19 22:04:39 gkar kernel: [Hardware Error]: Machine check events logged
Mar 19 22:11:40 gkar ntpd[2457]: synchronized to 149.20.68.17, stratum 2
Mar 19 22:21:07 gkar kernel: imklog 5.8.10, log source = /proc/kmsg started.
Mar 19 22:21:07 gkar rsyslogd: [origin software="rsyslogd" swVersion="5.8.10" x-pid="2435" x-info="http://www.rsyslog.com"] start
Mar 19 22:21:07 gkar kernel: Initializing cgroup subsys cpuset
Mar 19 22:21:07 gkar kernel: Initializing cgroup subsys cpu
Mar 19 22:21:07 gkar kernel: Linux version 2.6.32-358.2.1.el6.x86_64 (mockbuild@c6b8.bsys.dev.centos.org) (gcc version 4.4.7 20120313 (Red Hat 4.4.7-3) (GCC) ) #1 SMP Wed Mar 13 00:26:49 UTC 2013

I've disabled IPMI, loaded the w83795 ko and run voltcheck.sh multiple times. I notice a voltage drop on CPU3 compared to the other CPUs (even running at 3.4GHz). For example, when I had things set to 1.1875V, CPU3 was running 1150 mV while other CPUs were right at 1200mV. While running at 3.5GHz (stable overnight at 1.2Vcore), I see the following:

CPU1 (min/avg/max): 1204/1207.6/1212 mV
CPU2 (min/avg/max): 1186/1188.0/1190 mV
CPU3 (min/avg/max): 1168/1170.4/1174 mV
CPU4 (min/avg/max): 1192/1196.8/1204 mV

I'm not sure if this is related to http://hardforum.com/showthread.php?t=1750424 or not. I did confirm that node4/5 are actually CPU2 on the motherboard when running "tpc". Not sure what other utilities this affects.

Does it make sense to up the voltage on just CPU3 while trying 3.6GHz to see if it stabilizes? How do I match voltcheck's CPU3 to tpc nodes because of the discrepancy mentioned above? Thanks.

 

[plext0r]

I relocated the machine, installed mcelog and got a single MCE as follows (no panic this time):

Hardware event. This is not a software error.
MCE 0
CPU 54 BANK 5
MISC 0 ADDR 1e
TIME 1363963083 Fri Mar 22 10:38:03 2013
STATUS dc00000000020e0f MCGSTATUS 0
MCGCAP 107 APICID 66 SOCKETID 3
CPUID Vendor AMD Family 21 Model 1

This was while running at 3.6GHz with 1.2125V specified. I went back down to 3.5GHz at 1.2V and we'll see if it remains stable. Current voltcheck:

./voltcheck.sh
CPU1 (min/avg/max): 1206/1209.6/1214 mV
CPU2 (min/avg/max): 1192/1192.8/1196 mV
CPU3 (min/avg/max): 1170/1175.6/1184 mV
CPU4 (min/avg/max): 1196/1202.0/1206 mV