3960x dieing?

[tunatime]

Anyone know if ccxs/cores tend to die on zen?

My 3960 rig has stopped booting anything after bios unless I turn on core count 1+1 on my gigbite mb. Doesn't matter if it's a USB drive, ssd or m2. It just freezes on the aorus screen.... oddly it works in bios fine and gives no post error codes.

Checked all ram and the 8 cores it lets me boot with and it passes prime, mem test 12hours cinebench. Tried underclocking, resetting the CPU/block...it just will not boot anything with all cores/ccds on and I see no way in bios how to pick specific cores to test...

Does this sound like a MB or CPU? And what would be the best way to troubleshoot this I never had a CPU act up like this before

 

[Nobu]

Tried memtest86 with all the cores? If there's something wrong with a core, it'll probably show up there (assuming it even runs).

 

[noko]

One core on my 3960 became unstable anything beyond 4ghz, basically I turned off Core Performance Boost so none of the cores would boost and it would boot up just fine, run Windows etc. Yours sound very similar to my experience, same exact symptoms, mine was CPU.

Some things I did:

• Then I manually OC to see if I could find a point where it failed and that was around 4ghz
• I then manually OC one CCX at a time to over 4ghz, something like 4.2ghz and found the core causing the problem
• I then OC all the other CCXs, keeping 1 CCX slow
• Performance was affected weirdly, depending upon Windows scheduling, it could use the much slower core and basically tank single thread stuff
• I RMA it to AMD indicating my findings on which core went bad and they replaced it. They sent a new in box replacement which performs better than the first 3960.

Reason why it can boot into bios is most likely is that it is not boosting to failed point.

Why the first 3960x failed, I did not use PBO for any length of time except for testing, only using Ryzen Master, never in the bios until the core failed. Used Ryzen Auto OCing once again for testing but never used it beyond that. Since these CPUs are the bad of the bad, higher power, one less core per chiplet, there maybe more prone to failure and really I do not know. I have a thread here going through this, too lazy to find.

Edit: Found it:
https://hardforum.com/threads/what-else-can-go-wrong-in-2020-ryzen-tr-3960x-issues.2005678/

 

[tunatime]

One core on my 3960 became unstable anything beyond 4ghz, basically I turned off Core Performance Boost so none of the cores would boost and it would boot up just fine, run Windows etc. Yours sound very similar to my experience, same exact symptoms, mine was CPU.

Some things I did:

• Then I manually OC to see if I could find a point where it failed and that was around 4ghz
• I then manually OC one CCX at a time to over 4ghz, something like 4.2ghz and found the core causing the problem
• I then OC all the other CCXs, keeping 1 CCX slow
• Performance was affected weirdly, depending upon Windows scheduling, it could use the much slower core and basically tank single thread stuff
• I RMA it to AMD indicating my findings on which core went bad and they replaced it. They sent a new in box replacement which performs better than the first 3960.

Reason why it can boot into bios is most likely is that it is not boosting to failed point.

Why the first 3960x failed, I did not use PBO for any length of time except for testing, only using Ryzen Master, never in the bios until the core failed. Used Ryzen Auto OCing once again for testing but never used it beyond that. Since these CPUs are the bad of the bad, higher power, one less core per chiplet, there maybe more prone to failure and really I do not know. I have a thread here going through this, too lazy to find.

Edit: Found it:
https://hardforum.com/threads/what-else-can-go-wrong-in-2020-ryzen-tr-3960x-issues.2005678/

Well shit I changed to per ccx clocking at 1ghz and tried each one up to 4ghz and they worked.....time for more testing

Yours that failed how did it clock? Mine did 4.3 cinebench 4.55 light gaming

 

[noko]

Well shit I changed to per ccx clocking at 1ghz and tried each one up to 4ghz and they worked.....time for more testing

Yours that failed how did it clock? Mine did 4.3 cinebench 4.55 light gaming

Mine degraded rapidly after a few weeks of blue screening before it would not boot into Windows. I never did an overall OC of the cores prior to the issue, it was just PBO and Auto OC in Ryzen Master. New CPU has never been OC in bios or Ryzen Master. I just never found the previous OC to have any meaningful significance with much higher power/temperatures. AMD ended up replacing it with very little questions from them, I just laid down everything I did, gave images of bios settings that allowed it to run and which ccx had the issue. They will want the serial # of the CPU, picture of the CPU and purchase receipt.

 

[tunatime]

These things are odd I miss the days of it either works or doesn't. Also I'm not sure it's the chip I found a discoloration around one of the vrm caps ....but for now Upping loadline calibration and a .035 v core bump appears to work but it's only boosting to 4.3

does that look like one of the(supposedly solid) caps leaked to y'all? It doesn't look or smell burnt and held up to the let's poke it and see if it falls off test

 

[nealx01]

These things are odd I miss the days of it either works or doesn't. Also I'm not sure it's the chip I found a discoloration around one of the vrm caps ....but for now Upping loadline calibration and a .035 v core bump appears to work but it's only boosting to 4.3

does that look like one of the(supposedly solid) caps leaked to y'all? It doesn't look or smell burnt and held up to the let's poke it and see if it falls off test

View attachment 365920

its likely just oil from the VRM thermal pads. the middle of the VRM would in theory have the highest thermal density. i highly doubt thats any issue whatsoever