Question Memtest86+ errors

[O1111]

I've been getting stability issues lately (i.e. compilation of huge projects fails with random errors). Memtest86+ showed a few errors for memory slot B2 (I've tried both sticks from the kit and also a separate 8 GiB stick I bought just to test this), about 64 of them or so (8 here, but will be more if left for a bit longer):

In addition, I bought a new motherboard to test whether it is the problem and now I'm getting thousands of these errors each second in either slot:

Original configuration:
Motherboard: MSI X670E Carbon Wi-Fi
CPU: 7950X
Cooler: Artic Liquid Freezer II 420
RAM: 64GB DDR5 (Kingston Fury KF560C36BBEK2-64)
SSD: Samsung 990 Pro 2TB, 970 Evo Plus 1TB
PSU: DeepCool PX1000G
New motherboard (cheapest I could find with 2 M.2 slots): ASRock B650M-H/M.2+
CPU is the only thing (well, also the cooler) which was the same in all tests. Is this a dead CPU or still could be something else?
I'm quite sure I mounted it correctly as this is not the first time I'm installing a CPU into LGA socket, but 8700k worked without any issues previously.

 

[helpstar]

try latest memtest version 7.20
update the bios of your motherboard
clean the landing pads if the cpu with contact spray or isopropyl alcohol
be sure not to fasten the screws of the cooler too much, losen these a little and check memtest again
try the motherboard outside the case

 

[geofelt]

Even one error in memtest is cause for a rma.
Call Kingston customer support for a RMA; your ram should have a lifetime warranty.
Verify that your motherboard bios is current.
Many updates fix ram compatibility issues.

 

[O1111]

Even one error in memtest is cause for a rma.
Call Kingston customer support for a RMA; your ram should have a lifetime warranty.
Verify that your motherboard bios is current.
Many updates fix ram compatibility issues.

Yeah, but I'm not sure whether it's a CPU, motherboard or RAM fault.

I've decided to give the old motherboard another chance (because it has more IO and gives higher boost clocks), but I'm still getting these RAM errors. It's consistently happening on the second pass of memtest (test 2).
BIOS is the latest non-beta one on the "old" motherboard.

https://imgur.com/a/r2yMoCM

It's showing errors in the same exact bit quite consistently.

 

[Roland Of Gilead]

Yeah, but I'm not sure whether it's a CPU, motherboard or RAM fault.

Be sure. It's memtest86+. It's only testing memory. Your memory is failing. It has multiple errors. a bios upgrade or clear CMOS will not fix it. Sadly you you should RMA the ram, or prepare to get a new kit.

Test them one by one to be sure which is defective or if both are.

 

[O1111]

Be sure. It's memtest86+. It's only testing memory.

Technically, it doesn't test only memory. It could be a faulty RAM controller in the CPU or even something with the motherboard

Test them one by one to be sure which is defective or if both are.

That's the thing: I also tested them one by one (each stick was tested in each slot) and got similar results. I even tested another freshly bought 8 GiB stick and also got memory errors:

https://imgur.com/a/yvdJRhb

 

[geofelt]

Your tests with multiple motherboards and latest bios along with different ram suggests to me that there is a problem with your 7950x processor.
RMA time, I think.

 

[O1111]

It was only one test with a different motherboard, so I made another one. It's been 2.5 hours and it's running without any errors whatsoever. I think it's safe to consider the memory stable on the new motherboard.

So either a faulty old motherboard or I somehow installed the CPU incorrectly into the old one for like ten times straight but got it first try with the new motherboard. Or maybe kind of both, because the old motherboard somewhat worked for quite some time (though it's not the first time this happened, see my previous posts in the February). I guess I should RMA it.

 

[gamerk316]

At this point, I'd re-test on the original motherboard, making sure to double-check all the memory speeds/timings. You also might want to run a test at stock, in case the auto-OC profile is unstable for some reason.