I'd be a bit afraid about doing a fixed overclock like that. The 1.3V you use is fine when there's just a few cores are in use but the same voltage could be dangerous when all cores are at 100%. When the normal boost behavior of the CPU is in use, the CPU will ask for different voltages depending on its temperature and usage. It will be able to go down with the voltage before things get dangerous.
I found a screenshot about the BIOS settings I mean from a different MSI motherboard:
https://i.imgur.com/n6ii9nM.png
I mean use a setting "CPU core voltage" = "offset mode" and then "CPU offset mode mark" = "+" (plus) and then "CPU offset voltage" a small value like "0.01". Everything else about the BIOS settings would be kept at defaults, CPB would be kept enabled etc.
I did that. When I do offset with + and small amount if offset then cpu is locked at 3700 MHz
]]>I found a screenshot about the BIOS settings I mean from a different MSI motherboard:
https://i.imgur.com/n6ii9nM.png
I mean use a setting "CPU core voltage" = "offset mode" and then "CPU offset mode mark" = "+" (plus) and then "CPU offset voltage" a small value like "0.01". Everything else about the BIOS settings would be kept at defaults, CPB would be kept enabled etc.
]]>This shouldn't happen. It should still boost normally.
See if you misunderstood something about the BIOS settings, maybe there are different settings that can do something to the core voltage.
I checked bios so many times...
Maybe I did a mistake somewhere but at the end I'm pretty sure I did everything right.
There are 2 situations when system is stable
After core performance boost disabled (everything else auto)
Manual core multipler set. I checked 42 what gave me 4200MHz for all cores. But with this settings cpu even in idle doesn't go to it's idle values what is obvious.
I think that new bios settings could help but. Last bios from msi is from the middle of November and it's a beta.
Little Update
I checked manual settings.
Multiplier 43, vcore 1.3V
It seems stable. Blender benchmark passed with the best result yet.
See if you misunderstood something about the BIOS settings, maybe there are different settings that can do something to the core voltage.
]]>"WHEA" error/warning events in Windows are the same as the "MCE" log entries in Linux.
]]>That "mce" error in your log from earlier is (usually) a mistake that happens inside the CPU. Did you try playing around with the Vcore voltage setting in the BIOS to see if the behavior changes?
If you can't fix the problem you have to return either the motherboard or CPU and ask for a replacement. You can try asking the motherboard manufacturer for help before you do this. Do you currently use the newest BIOS version for your board? You could also see if the manufacturer offers a beta version for your board and try that.
I use latest bios and it is beta. I checked previous one too.
I'm not sure how to adjust voltages. I know where change voltage values but not sure about values. (Searched but there are different opinions)
I was advised to adjust ddr voltages and SoC voltages hut it didn't help.
Just a minute ago another reboot occurred in Witcher 3.
After grub there was an information:
[0.148839] mce: [Hardware Error]: CPU17: Machine Check: 0 Bank 1: bc8008000 60c0859
[0.148839} mce: [Hardware Error]: TSC 0 ADDR 1b39382c0 MISC d012000000000000 IPID 100b000000000
[0.148839] mce: [Hardware Error]: PROCESSOR 2:a20f10 TIME 1609012524 SOCKET 0 APIC b microcode a201009
(I took a picture of the screen with my phone)
There are a lot of people with reboot problem but its windows error
WHEA-Logger ID 18 and critical error Kernel-P
Don't know if this is the same problem but it seems that I'm not alone
]]>If you can't fix the problem you have to return either the motherboard or CPU and ask for a replacement. You can try asking the motherboard manufacturer for help before you do this. Do you currently use the newest BIOS version for your board? You could also see if the manufacturer offers a beta version for your board and try that.
]]>https://developer.blender.org/T73648
Seems like the particular test is pretty RAM intense.
Since you ruled out temperature from CPU & GPU (and thus likely undercurrent), the problem is most likely in that area.
Did you replace all RAM modules? Did you attempt the most conservative RAM timings (BIOS setting)?
Yes. I replaced all modules
2x16GB hyperx replaced with 2x16 G.skill
Different brand, speed, timings.
Both sets of ram I checked on default settings and with xmp profile on.
After replacing memory I did 3 blender benchmark tests and run Wither 3 (it was another application causing reboots but usually after 2 or 3 minutes) for about 20-30 minutes.
I though that gpu is not used in this test since st the beginning I had to choose gpu or cpu for test.
I have this gpu since 2013 or something. With previous config there was never a problem and I used it for all my blender renders
Max temp for gpu I remember was around 70 or littlemore degrees so not bad i think.
I didn't have time to run witcher again.
I don't use swap
Can this cause problems like this with 32GB of ram ?
Seems like the particular test is pretty RAM intense.
Since you ruled out temperature from CPU & GPU (and thus likely undercurrent), the problem is most likely in that area.
Did you replace all RAM modules? Did you attempt the most conservative RAM timings (BIOS setting)?
But it's quite weird situation.
Pc rebooted on blender benchmark test 6. Its last test. It's called Victor.
I wanted to try something else.
I run mprime. Test 4 - mixed
I'm not sure how long but much longer thannkende. I had to turn off pc but it didn't reboot. Mprime can cause cpu to he really hot. 90 degrees at max but there was no reboot.
Max temp in blender was 75 so much lower.
How can I save a log especially for blender benchmark?
]]>Sounds to me like the CPU is not getting the voltage it needs.
Can u disable Precision Boost 2 and test again?
Ok
There is no PB2, but I've found option CPB
Core Performance Boost
It seems that when it is disabled pc is stable.
But it limits core locks
Default auto boost from 3.7 to 4.x (not sure)
When CBP is disabled results in blender benchmark are about 25 % worse
Edit.
I did bios reset (hardware forum advice). Didn't help but during blender test I saw something.
During all tests (in blender benchmark there are 6 tests) fans were loud (fast rotation)
In 6th test fans slowed down to standard speed and after about 2 second PC rebooted.
I think that It can be hardware bios problem. I'm searching fo the solution and it seems that there are a lot people having problems with 5900x and 5950x
]]>