Random Non-Load related Computer Lockups

[magibeg]

Are you using new prime95 with AVX or old prime95 ? old prime isn't very good at breaking modern CPU's. p95v296b7.win64.zip <latest prime95 with AVX
Are you using the Samsung NVME driver ? I assume you are since Magician says everything is OK.
I would use something to generate a sensor log so you can see what the temps voltages and fan speeds where before the crash, it may be something as stupid as gpu fans getting stuck in idle off mode or a weird quirk with CPU power states that results in a crash.

Additionally are you using any other devices with the system not listed in specs such as WIFI/BT adapters, sound cards, USB hubs, webcams that sort of thing?

I'm using the new prime95 with AVX (damn you AVX for making overclocking more complicated!)

Magician says everything is perfect.

A sensor log actually sounds like a good idea, do you have any recommendations?

As for additional items, i don't have anything non-essential. I unplugged all devices I could when I suspected it was my rocksmith USB cable.

 

[Folterknecht]

A sensor log actually sounds like a good idea, do you have any recommendations?

HWinfo64 has a logging function exporting in .txt files I believe.

 

[er557]

I would highly recommend, due to possibly a sound card or pcie issue you are experiencing, dpi latency and such,
to enable hpet in bios, and force it in windows, also the change to hardware timers might alter your situation for the better. plus a more smooth experience all around.
use these commands in admin prompt and reboot:

bcdedit /set useplatformclock yes
bcdedit /set tscsyncpolicy Enhanced
bcdedit /set disabledynamictick yes

 

[Deleted member 67555]

I thought regardless of bios settings HPET was no longer used in windows 10 1803 and later.

 

[er557]

I will look into this, although i am pretty sure you can force it on in any windows version, and the evidence for that is otherwise x265 hwbot benchmark will not run if not enabled.
It might also settle down a bit the cpu peak performance, but worth it for the smooth experience all around.

 

[FCG]

@magibeg
Are you making use of Load Line Calibration or similar feature to modify Vdroop from the standard setting?

 

[magibeg]

@magibeg
Are you making use of Load Line Calibration or similar feature to modify Vdroop from the standard setting?

Nope, though when I was initially overclocking there was little to no Vdroop at all under load.

I'm running HWinfo64 in logging mode now, it's going to spit out colossal amounts of data it looks like. Now i just need to wait for a crash to happen again...

Alright I actually just had a crash with the HWinfo64 log running.

Screen froze up completely, however music from the youtube video i was watching kept playing for a few minutes, though the whole computer was completely unresponsive.

Going through the event log the only thing that sticks out is a pair of errors about nvlddmkm stopped responding following by a warning claiming that it recovered (it's a display driver). Though that log entry hasn't been present in any of the previous crashes.

The log file is pretty massive that HWinfo makes so it'll take a bit to poke through it.

In case anyone would like to look at the log here it is:

http://s000.tinyupload.com/index.php?file_id=02090260877369278309

So the final entry matches the time stamp of the nvlddmkm errors. So that sort of implies a video card issue. Something of note for the logs is that everything looks normal except the last entry has my CPU maxed out in clock speed (running stock). Didn't see anything else that would indicate odd behavior.