[SOLVED] Seasonic M12II 620W Dying?

Status
Not open for further replies.

quadrax

Honorable
Feb 5, 2018
117
7
10,585
Hello all,

My problem is (probably) with my Seasonic M12II 620 Evo PSU.

Specs of PC.

MB : ASUS-TUF-Z490 gaming.
CPU : Intel Core i5 10600K (overclocked all cores to 4.7GHz @ 1.3v)
GPU : EVGA RTX 3070
RAM : 4x16 GB RAM DDR4 3200MHz (OCed to 3300 MHz)
HDD : 2x SATA HDD (7.2k RPM)
SSD : 2x M.2 NVMe SSD (1TB + 500GB Samsung Evo Plus)
CPU Cooler : Noctua NH-D12
OS : Windows 11 Pro - Legal

Problem : When i playing games i gotta sometimes error (no matter which game). So i decided and tried undervolt GPU = same result, so i revert back to stock frequency of GPU (for testing) and problems occurs again. So i tried one more thing and that was "burn" CPU with highest possible workload (AVX instructios) over Y-Cruncher, and then playing game which is extremely GPU demanding. (100% gpu usage all the time, because VSYNC off and 200% Scale resolution). And about 5 min later i got BSOD.

So my thoughts are - PSU dying ? PSU is about 2 years old. If iam correctly count - 160W Ycruncher, 255W GPU. It should be ok, but seems like not.

Any ideas what to do? I cant try other PSU nor i dont have other one. In normal case (Web browsing etc. light loads is ok)

I honestly fear, that my PSU dying, slowly, but it does.
 
Solution
Just for info, i ended with 1.330v for 4.1GHz. Do you guys think it still could be a PSU problem? Or CPU degraded so fast and so much. Thanks
Both are probable. A noisy PSU creates more work for the CPU and GPU VRMs, which could be enough to make barely-working OCs fail. Silicon degradation can cause barely-working OCs to fail over time too. When I experimented to find out how much I could under-volt my i5-11400 by, it worked down to -70mV offset, so I settled for -50mV to avoid being too close to barely-working. Same for my memory under-volting, worked down to 1.17V from 1.35V in XMP, so I rounded it up to 1.2V.

quadrax

Honorable
Feb 5, 2018
117
7
10,585
the 3060TI requires a little less power than the 3070 and that extra the latter needs may be pushing the PSU over the limit. If it is not the PSU it would be the GPU so can you test it in a different system?
I dont have other PC to test it. But i just retry again - Y-Cruncher + playing game (legal of course). And problem occurs again. I just monitored temperatures and CPUs are with testing about mid 70, but i saw a when it was under fully load (before freeze, not even hard reset works), on 12V rail was only 11.808V (over HWInfo) . Could be problem there?

Even if i tried other game, same problem.
 

quadrax

Honorable
Feb 5, 2018
117
7
10,585
I just figured out where the problem was (probably). When i had opened HWInfo, so i check WHEA errors, and there was L0 Cache and CPU, but what i dont understand is WHY. Everything was ok, before i installed 3070. I still suspect PSU is dying. So i just added little more juice to CPU and seems stable atm. But in long terms i would like to get it back. Maybe with new PSU i can do it. Hardly to believe to CPU degradation so fast. Overclock was not even MID-range, just 4.7GHz @ 1.3v. Now i have 1.315v
 

InvalidError

Titan
Moderator
HWInfo / SMB voltage monitoring is rarely accurate as platform monitoring chips have loose tolerances. Use a real multi-meter if you want to know what output voltages actually are.

While it is possible for a noisy PSU output to cause crashes, dying PSUs are more likely to cause an unexpected shutdown from tripping its own protections or reset from something else panicking and pulling the reset line to prevent (further) data corruption.
 

quadrax

Honorable
Feb 5, 2018
117
7
10,585
HWInfo / SMB voltage monitoring is rarely accurate as platform monitoring chips have loose tolerances. Use a real multi-meter if you want to know what output voltages actually are.

While it is possible for a noisy PSU output to cause crashes, dying PSUs are more likely to cause an unexpected shutdown from tripping its own protections or reset from something else panicking and pulling the reset line to prevent (further) data corruption.

Well. It's maybe a beginning of failing (if it does), but i am gonna buy new PSU unit anyway. So it should be then ok. Going for Seasonic GX 1000w Gold.
 

quadrax

Honorable
Feb 5, 2018
117
7
10,585
So i thought problem is gone with raising vCore, but no. It's same. So i put default core ratio 4.1GHz and 1.325v for testing now. Lets see if LinX show me again error... If yes, then its PSU or MB.

EDIT : Even on default i am getting errors. Any ideas what now?
 

InvalidError

Titan
Moderator
Just for info, i ended with 1.330v for 4.1GHz. Do you guys think it still could be a PSU problem? Or CPU degraded so fast and so much. Thanks
Both are probable. A noisy PSU creates more work for the CPU and GPU VRMs, which could be enough to make barely-working OCs fail. Silicon degradation can cause barely-working OCs to fail over time too. When I experimented to find out how much I could under-volt my i5-11400 by, it worked down to -70mV offset, so I settled for -50mV to avoid being too close to barely-working. Same for my memory under-volting, worked down to 1.17V from 1.35V in XMP, so I rounded it up to 1.2V.
 
Solution
Both links support that it could be the PSU which struggles at high loads when released it was a budget line to compete with the likes of Corsair CX
 

quadrax

Honorable
Feb 5, 2018
117
7
10,585
Both links support that it could be the PSU which struggles at high loads when released it was a budget line to compete with the likes of Corsair CX

Thank you for these links. Useful informations are there.
 

quadrax

Honorable
Feb 5, 2018
117
7
10,585
Just for information PSU really dying. I removed GPU, and CPU tests now pass without any problem (even on my rock stable OC 4.7GHz @ 1.3v). So rip my PSU, gonna get other one.
 
Status
Not open for further replies.