My (fairly high-end) PC just insta-shuts down when I run my own python based AI video processing software (fairly high load, not sure if it uses much GPU at all - not set to, the AI models are the CPU ones, lots of memory related operations, copying frame data around, writing frames to disc at same time etc.). But only happens on one particular video clip - all other clips run without problems... so something in this particular clip does it. This does not happen for any other software on this PC!
As others mentioned - the only way to get back on is to physically switch the power on back of PSU off to on again.
I have read a lot of answers here and it all appears to say the ONLY device in a high-end PC which could cause this kind of instantaneous power off without any normal "bluescreen" is the power supply itself tripping some protection circuit, is that correct?
This is not a low end setup - although the PSU is a bit under-powered... but no problems up to now
I was not running any overclocking at all on GPU or CPU... the only thing that could be considered an "overclock" was running the RAM at the DOCP profile provided by the AM manufacturer that it is supposed to be specified for.
The problem was 100% reproducible running my software on this clip.
I tried turning off all third party software which might affect CPU (like Lasso) - still happened
Tried turning off that "Resizeable BAR" for GPU in the BIOS - still happened
Then I decided to turn off DOCP and run the memory at default BIOS setting - this FIXED IT - I could run my software on the problem clip and it completed without failing
So question is... how could a DOCP profile on memory certified by the maker cause a PSU to power trip protect itself?? I only recently bought this RAM after upgrading from previous 32GB setup.
Could this be some Ryzen problem with running 3600Mhz on the CPU memory controller fabric (it would never run this on the old RAM, but seems to be stable on this new made for Ryzen DOCP RAM at 3600Mhz) - don't know how this could trip a PSU though.
I did use latest (v12.1.11 registered to me) OCCT software to monitor the machine during a run to see if I saw any crazy power or temperature usage, and I didn't see any up to point of instant power off. I don't think the actual problem would have time to display or save somewhere before power goes.
Just looking for any other ideas before I go to expense of buying a larger PSU - don't want to buy one and then find problem still there!
Specs:
Click banner image link in signature below to get full specs of current running state on CPU-Z site
Motherboard: ASUS Hero VIII X570 - all BIOS set to stock settings - currently running 4402 BIOS version (have not upgraded to 4702 yet)
CPU: Ryzen 5950X stock settings, no overclock
RAM: 64GB G.Skill TridentZ Neo 3600 (32GB x 2) (specs on box - CL18-22-22-42 1.35V "Designed for Ryzen")
GPU: Gigabyte 4080 Gaming OC (running stock settings), GPU vertical mount using Coolermaster V3 mount
PSU: EVGA 750 G3 Supernova GOLD
Cooling: CPU - EK AIO 360 Elite RGB (6 fan push pull) - case another 8 fans
Case: HAF 700 EVO
As others mentioned - the only way to get back on is to physically switch the power on back of PSU off to on again.
I have read a lot of answers here and it all appears to say the ONLY device in a high-end PC which could cause this kind of instantaneous power off without any normal "bluescreen" is the power supply itself tripping some protection circuit, is that correct?
This is not a low end setup - although the PSU is a bit under-powered... but no problems up to now
I was not running any overclocking at all on GPU or CPU... the only thing that could be considered an "overclock" was running the RAM at the DOCP profile provided by the AM manufacturer that it is supposed to be specified for.
The problem was 100% reproducible running my software on this clip.
I tried turning off all third party software which might affect CPU (like Lasso) - still happened
Tried turning off that "Resizeable BAR" for GPU in the BIOS - still happened
Then I decided to turn off DOCP and run the memory at default BIOS setting - this FIXED IT - I could run my software on the problem clip and it completed without failing
So question is... how could a DOCP profile on memory certified by the maker cause a PSU to power trip protect itself?? I only recently bought this RAM after upgrading from previous 32GB setup.
Could this be some Ryzen problem with running 3600Mhz on the CPU memory controller fabric (it would never run this on the old RAM, but seems to be stable on this new made for Ryzen DOCP RAM at 3600Mhz) - don't know how this could trip a PSU though.
I did use latest (v12.1.11 registered to me) OCCT software to monitor the machine during a run to see if I saw any crazy power or temperature usage, and I didn't see any up to point of instant power off. I don't think the actual problem would have time to display or save somewhere before power goes.
Just looking for any other ideas before I go to expense of buying a larger PSU - don't want to buy one and then find problem still there!
Specs:
Click banner image link in signature below to get full specs of current running state on CPU-Z site
Motherboard: ASUS Hero VIII X570 - all BIOS set to stock settings - currently running 4402 BIOS version (have not upgraded to 4702 yet)
CPU: Ryzen 5950X stock settings, no overclock
RAM: 64GB G.Skill TridentZ Neo 3600 (32GB x 2) (specs on box - CL18-22-22-42 1.35V "Designed for Ryzen")
GPU: Gigabyte 4080 Gaming OC (running stock settings), GPU vertical mount using Coolermaster V3 mount
PSU: EVGA 750 G3 Supernova GOLD
Cooling: CPU - EK AIO 360 Elite RGB (6 fan push pull) - case another 8 fans
Case: HAF 700 EVO
Last edited: