Different Versions of Same RAM Causing Instability?

[MCDELTAT]

Is it possible for two different sets/versions of RAM to cause instability even if they are the same brand, CAS Latency, Everything?

 

I'm spare some of the details, but my Linux rig recently started getting randomly hung up. It would function fine, no issue, then it full hangs where the Networking stack dies, GUI/Compositor Dies, everything. Reading through the errors online made me take a guess at my RAM. I made a bootable thumb drive of memtest86 and ran it on all 4 DIMMs that were installed in my system. I got 44 errors per loop.

 

So I just took the time to put them in individually and run through 2 loops of the same test to find the culprit. On their own, none failed, 0 errors. Is that possible? Are weird micro issues the culprit? Am I allowed to return them in hopes of getting a matching pair?

 

RAM:

64GB (4x16GB) Corsair Vengeance LPX DDR4 3600MHz CL 18-22-22-42

Difference: 2 Stick were purchased in May 2020 and are Ver 3.44. The other 2 were purchased July 2020 and are Ver 5.49

XMP Profiles are stock as listed above, never bothered overclocking or changing any timings.

[Hairless Monkey Boy]

It is possible that they will not work together, especially since you bought them at different times and they are different versions. Brands can change the PCB and ICs, so that subtiming requirements may be different even though the primary timings are the same. It's also possible that the XMP profile works with 2 sticks, but not 4, even if they were identical sticks.

 

There are a few things you can try to get them working, but first you should verify that each kit works individually (you have tested individual sticks, but not individual kits).

 

Once you verify that that is all working, try them together again, but try these orientations:

ll ll

ll ll

ll ll

where green is one kit and blue is the other. The most likely to work is the last since matching sticks will be in their own channels.

[SupaKomputa]

Yeah it's possible.

Check the ram with Memtest86.

Also depending on the motherboard / cpu, they don't like the ram slot fully occupied, you might wanna run them at lower speed.

[MCDELTAT]

1 hour ago, HairlessMonkeyBoy said:

There are a few things you can try to get them working, but first you should verify that each kit works individually (you have tested individual sticks, but not individual kits).

Good point. I'll test each kit next in memtest86. I think I had either Pattern 1 or 2 from your diagram because I installed the first kit in alternating slots and then just added the new kit when I got it a few months later.

 

1 hour ago, SupaKomputa said:

Also depending on the motherboard / cpu, they don't like the ram slot fully occupied, you might wanna run them at lower speed.

For the sake of internet history, I'm currently running these on:
Intel 8700K
Asus Prime Z370-A

BIOS Version 2401

Asus 3070 Dual Driver 460.73.1 (Thanks LTT for this in the first drop).

[SupaKomputa]

Not an expert on intel platform, but the full slot lower speed is a thing in AMD, and 8700k natively support up to 2666mhz.

I suggest reducing the speed to 3200 or 3000 if you wanna get stable with all slot.

 

[MCDELTAT]

Update:
Ran memtest86 on 2 DIMMs, one from each pair. No errors. So It might honestly be just when it's all 4 DIMMs. Next test is all 4 as I think they were (alternating blue/green) and then the versions in their own channels. Very interesting...