Problem with system for DC

[Markfw]

OK, so I have an MSI something motherboard, a 7950x and a 4090 (main components). Normally it runs 100% CPU load and 100% GPU load (WCG + F@H). Works fine for months. So its over 100F temp here, and my AC can't handle all the computers, so I turn off (pause) the main heat producer, several of the 4090 cards. When I do, this one locks up in an hour or less. But fully loaded, no problem. PSU is EVGA 850 G5 gold rated.

To reiterate, loaded is fine, video idle, locks up !

Ideas ???

 

[craiggloyd]

Markfw i see you on the stats page of WCG awesome setup. Just 1 of your identical machines? I would just try KISS, reduce the CPU turbos a bit or try 105W efficiency mode just to see if you can establish simple stability. The way you have it setup, if Folding & WCG do you have cores dedicated (set Affinity) to Folding-GPU so some cores not fully utilized but once you pause that, all cores become fully loaded if you also modify affinity of BOINC to use every core meow? Or are all CPU cores maxed in both scenarios? Different workloads and something on the edge it locks up?
From what I've seen on my Intel's, Rosetta load the cores harder than WCG (same watts, less clocks)

 

[Markfw]

Markfw i see you on the stats page of WCG awesome setup. Just 1 of your identical machines? I would just try KISS, reduce the CPU turbos a bit or try 105W efficiency mode just to see if you can establish simple stability. The way you have it setup, if Folding & WCG do you have cores dedicated (set Affinity) to Folding-GPU so some cores not fully utilized but once you pause that, all cores become fully loaded if you also modify affinity of BOINC to use every core meow? Or are all CPU cores maxed in both scenarios? Different workloads and something on the edge it locks up?
From what I've seen on my Intel's, Rosetta load the cores harder than WCG (same watts, less clocks)

Rosetta is unstable on most of my 7950x's, no idea why, so they only do WCG. But its odd, as I said the more load, the better, and it locks up with no GPU load. So lightening the CPU load would only get worse.

 

[craiggloyd]

Rosetta is unstable on most of my 7950x's, no idea why, so they only do WCG. But its odd, as I said the more load, the better, and it locks up with no GPU load. So lightening the CPU load would only get worse.

Rosetta seems to load the units in the cores harder than other tasks besides probably Prime95 so perhaps it is taking it over the edge. Way less load maybe some cores turbo higher and causes a crash that way too. Efficiency mode I have heard is reasonable at slightly reducing clocks but cutting power significantly

 

[TennesseeTony]

GPU driver?

What is your power limit on the GPU? Maybe bump up the power limit (although I don't think power limit affects idle speed/power)?

 

[cellarnoise]

Are you still using a -20 or greater in the bios with curve optimizer? The lighter load may cause some cpu instability at that CO setting? Maybe try all core -8 co or even -5. Might help? And with the GPU idled are you still running the Boinc tasks with one core idled for feeding the now idled GPU?

 

[In2Photos]

Are you still using a -20 or greater in the bios with curve optimizer? The lighter load may cause some cpu instability at that CO setting? Maybe try all core -8 co or even -5. Might help? And with the GPU idled are you still running the Boinc tasks with one core idled for feeding the now idled GPU?

This was my first thought as well.

 

[Markfw]

This was my first thought as well.

default bios settings. MSI motherboard.

 

[TennesseeTony]

Switch PSU with another system and see if it persists. Or at least make sure the PSU fan is switched to ALWAYS ON.

 

[Markfw]

Switch PSU with another system and see if it persists. Or at least make sure the PSU fan is switched to ALWAYS ON.

PSU is an EVGA 850 G5, and they are one of the best, and it was brand new a year ago. Pretty core its the flake MSI motherboard. It will be replaced soon. Every MSI I have bought in the last year has had issues. Never again will I Buy them. ASRock is the only one right now. ASUS has been problematic, and I will never guy Gigabyte ever again, due to their RMA when lying about it.

 

[mmonnin03]

FTC notifies ASRock, Gigabyte and Zotac for warranty claims

Zotax exposes customer data

MSI had years of warranty claim data available to public in data leak
https://www.youtube.com/watch?v=DeE_gLl3j94&t=703s
(All from past week at GN)

ASUS makes it difficult to get free repairs under warranty (scams) Part 4.
https://www.youtube.com/watch?v=Z0ZoCYXmF0Q

They're all bad. All within the past 1 month