[RemoteSpecialist]

While I am waiting for another 5950x to be available - I'm taking a look for motherboards also.
Currently, I have MSI B550 Carbon WiFi - I cannot say a bad word about it. But what do think, guys? Is it enough for 5950x or I should consider smth else?

 

[danny9428]

Wait, you were getting reboots even with manual overclock? Were sleep states enabled?

Don't think I've ever tried disabling it when in manual oc
I can see Ryzen Master would report core going sleep but HWinfo64 doesn't show any C6 states
Maybe I can try and do that, see if that would change any while I wait for AMD to reply

It's just as frustrating as my old 6950X as this kind of reboot only ever happens when my system uptime is past 24 hours lol

 

[JohnnyFlash]

Don't think I've ever tried disabling it when in manual oc
I can see Ryzen Master would report core going sleep but HWinfo64 doesn't show any C6 states
Maybe I can try and do that, see if that would change any while I wait for AMD to reply

It's just as frustrating as my old 6950X as this kind of reboot only ever happens when my system uptime is past 24 hours lol

Ya, this is really surprising as most of my reading has shown that manual all core settings were fine. That's how I planned on running system, but now maybe I'll just get a 10980XE and sell my dark hero.

 

[danny9428]

Ya, this is really surprising as most of my reading has shown that manual all core settings were fine. That's how I planned on running system, but now maybe I'll just get a 10980XE and sell my dark hero.

If AMD lets me in through the RMA I might consider just go for the sTRX4 3960X instead
I should've just went HEDT but the Zen 3 just promised so much it's so tempting : P

 

[azomiss]

More details regarding PSIC setting.

3.In the BIOS, locate the Power Supply Idle Control option and set it to Typical (this option should be available in the Advanced section of the BIOS).
--- after changing this setting and booted in windows, system restarted in less than 5 minutes .
--- tried again and system got stuck for 30 seconds with a black screen and noticed that my USB keyboard did not have power.
--- system restarted by itself and bios has been automatically changed to default.

Also pictures with PSIC on auto and IDLE:

And with PSIC on Tyoical and IDLE:

I will get back with the Positive Curve Optimizer results also later today.

 

[GRABibus]

One final difference I'd like to point out between my original bad 5950x and the replacement - I consistently had a 7 to 8 degree C difference between CCD 0 and CCD 1.

The new 5950x has a 1-2 degree temperature difference between CCD's. MUCH more consistent and better matched.
Interestingly, it was CCD 1 (the cooler running CCD) on my old 5950x that had the most WHEA causing APICs.

From my side I have a 5900X since 1 month and half with no issue at idle or low loads.
CCX0 has 10 degrees more than CCX1 and my APIC whea’s at idle or low loads when tweaking PBO/CO (negative offset) come from Core8 (CCX1)....

So difficult to conclude....

 

[ghiga_andrei]

More details regarding PSIC setting.

3.In the BIOS, locate the Power Supply Idle Control option and set it to Typical (this option should be available in the Advanced section of the BIOS).
--- after changing this setting and booted in windows, system restarted in less than 5 minutes .
--- tried again and system got stuck for 30 seconds with a black screen and noticed that my USB keyboard did not have power.
--- system restarted by itself and bios has been automatically changed to default.

Also pictures with PSIC on auto and IDLE:

View attachment 2479996

And with PSIC on Tyoical and IDLE:

View attachment 2479997

I will get back with the Positive Curve Optimizer results also later today.

These BIOS settings are very sticky, once you change something from Auto to a specific value and you put it back to Auto it will remain on the last specific value. The only way to really test Auto is to clear the BIOS by CMOS battery removal. Your pictures are identical from my point of view, so they were both on Typical I would bet.

Don't forget to also extract the WHEA logs like I mentioned in the previous post. From Ryzen Master your best core is 02 so I would expect it to cause the faults since it will boost the highest. If your WHEA APIC IDs are always 2 or 3 then they relate to that core. Core 02 in Ryzen Master is actually core 1 in BIOS / Windows and the APIC IDs are relating to Threads and there are 2 threads per core so APIC ID 2 or 3 are the threads of core 1.

 

[ghiga_andrei]

From my side I have a 5900X since 1 month and half with no issue at idle or low loads.
CCX0 has 10 degrees more than CCX1 and my APIC whea’s at idle or low loads when tweaking PBO/CO (negative offset) come from Core8 (CCX1)....

So difficult to conclude....

I read somewhere on the AMD forum that if you have such a large temp difference between the CCDs (you say CCX but you mean CCD, CCXs are inside the CCDs) you can try to slowly unscrew one side of the cooler or the other one to see if it makes any difference. The theory would be that convex / concave metal casing of the CPU combined with concave / convex cooler base could cause it to be more tight on one CCD and more loose on the other. But it's just a theory.

My 1st 5900x also had a 15dgr difference between CCDs (CCD1 was hotter) and my 2nd 5900x has a 5dgr difference, but this time on CCD2. With the same cooler and I tried multiple installations and the results are pretty much the same.

 

[danny9428]

I read somewhere on the AMD forum that if you have such a large temp difference between the CCDs (you say CCX but you mean CCD, CCXs are inside the CCDs) you can try to slowly unscrew one side of the cooler or the other one to see if it makes any difference. The theory would be that convex / concave metal casing of the CPU combined with concave / convex cooler base could cause it to be more tight on one CCD and more loose on the other. But it's just a theory.

My 1st 5900x also had a 15dgr difference between CCDs (CCD1 was hotter) and my 2nd 5900x has a 5dgr difference, but this time on CCD2. With the same cooler and I tried multiple installations and the results are pretty much the same.

My 5950X has CCD0 running about 13~15 degrees hotter than CCD1
Haven't really checked much on how the IceGiant cooler pressures the chip though so I can't tell if doing the loose-screw offseting would help much
I don't feel like doing that on a 2kg weight air cooler though

 

[danny9428]

Ok so I just got a reply from AMD asking me to provide info on my CPU purchase and photo evidence for my RMA request
is it like you just reply to that generic support email or is it like there's a ticket website like Intel does?

It's my first time dealing with AMD regarding that so ... : P

 

[mtavel]

Ok so I just got a reply from AMD asking me to provide info on my CPU purchase and photo evidence for my RMA request
is it like you just reply to that generic support email or is it like there's a ticket website like Intel does?

It's my first time dealing with AMD regarding that so ... : P

I'm in the US and I just replied to their email. I never viewed/edited my ticket through any portal. It's possible they have different processes in different regions, but it definitely seemed like the agent I was corresponding with was in a very different time zone compared with me.

 

[azomiss]

These BIOS settings are very sticky, once you change something from Auto to a specific value and you put it back to Auto it will remain on the last specific value. The only way to really test Auto is to clear the BIOS by CMOS battery removal. Your pictures are identical from my point of view, so they were both on Typical I would bet.

Don't forget to also extract the WHEA logs like I mentioned in the previous post. From Ryzen Master your best core is 02 so I would expect it to cause the faults since it will boost the highest. If your WHEA APIC IDs are always 2 or 3 then they relate to that core. Core 02 in Ryzen Master is actually core 1 in BIOS / Windows and the APIC IDs are relating to Threads and there are 2 threads per core so APIC ID 2 or 3 are the threads of core 1.

Regarding WHEA-Logger .... 4,26,6,8,2,3,25,16

 

[ghiga_andrei]

Regarding WHEA-Logger .... 4,26,6,8,2,3,25,16

Seems something weird in your setup, cannot be so many cores that fail.
Please try also the curve optimizer. You may not have a bad CPU after all.

 

[GRABibus]

Seems something weird in your setup, cannot be so many cores that fail.
Please try also the curve optimizer. You may not have a bad CPU after all.

Curve optimiser with positive offsets of course 😊

 

[silot]

I have successfully RMA my 5900x , i retested and i am still getting WHEA errors with stock settings so the problem must be my GPU.

 

[GRABibus]

I have successfully RMA my 5900x , i retested and i am still getting WHEA errors with stock settings so the problem must be my GPU.

Or a second bad CPU ?
Do you have tried positive offsets with CO ?

 

[azomiss]

Okay, here is with PBO positive 10, now what?

 

[RemoteSpecialist]

I have successfully RMA my 5900x , i retested and i am still getting WHEA errors with stock settings so the problem must be my GPU.

What’s the batch number of a new CPU?

 

[ghiga_andrei]

Okay, here is with PBO positive 10, now what?

View attachment 2480037

Do you still get reboots with positive 10 ?

 

[MikeS3000]

So I finally received my new 5900x under warranty from AMD. That was fun. The package was stuck in Memphis with FedEx for 9 days before it finally got to me. My last CPU had the issue where it would fail Prime95 and OCCT at stock settings on the #1 core. My first impression of the new CPU (2104 PGS batch) is that the boost clock is toned down a bit. My old 5900x would try to boost just over 4.8 ghz running prime95 non-avx single core on the best cores. Now, the best cores are just shy of 4.8. My gut feeling is that some these failing CPUs were programmed to boost a bit too high from the factory and this led to failures at stock. I have lots of testing to do but so far no glaring issues.