Memory training issue using a TRX40 Creator motherboard

Hardware Hub Build a PC
,
1

Hey everyone. Had a look around to see if I could find my issue. This is the closest I could come to; Threadripper 9060x memory controller fault, 3 out of 4 sticks usable - SOLVED

I have an old ASRock TRX40 Motherboard with a Threadripper 3960X, 256GB of DDR4 3200, with 4 AI 9700.

256GB of RAM is 32GB DDR4 3200 MHz UDIMM PC4-25600 (PC4-3200AA) CL22 DIMM 2Rx8 Non-ECC

My problem;

B_1 and B_2 both show 0 in the bios and have memory training issues currently.

What I’ve done;

At first it showed A_1 A_2 B_1 B_2 as having memory training issues. So I researched it and found that it could be due to the threadripper’s seating.

So I carefully removed and re-seated the CPU, while also checking for any issues with the cpu slot. I couldn’t find any bends or anything that looked out of place. Everything looked clean and good inside of the slot itself.

After reseating and adding the cooler back, I gained A_1 and A_2.

I tried also tried what’s listed here; Checking your browser

Bringing the speed down to the absolute minimum and changing the voltage to 1.2 from auto, then 1.35 with no luck on those changes.

I also tested the RAM in different slots, with the same issue. The RAM in B_1 and B_2 originally work fine in A_1 and A_2.

What I need help with;

Without stating the obvious haha, but really!

  • Do I need to do things in a specific order for this motherboard/memory training issue?

  • I re-seated the CPU, should I continue to try to re-seat

I would appreciate any hints/tips on best practices please.

1 Like
2

Is this memory in the board’s QVL?

Anytime you use any configuration outside the QVL, here be dragons. In my anecdotal experience, SP3 family of platforms are especially picky about this.

Can you attempt to use this memory without XMP or any kind of OC? if the system works with memory locked to 2133, then there’s hope that some config can get stable at 3200.

Rule out PSU limitation?

1 Like
3

No unfortunately it’s not in the boards QVL, good point on that.

Yeah I tried it on auto settings or locked to 1.2v which is the memory default. I’ll pull all of the memory and cards and try just one slot at a time at lower speeds.

I haven’t had any other power related issues re: the PSU during load of the cards when running tests, or any random shut downs, etc. Other than not training the two slots, the system seems "stable” as in running without the RAM in those two slots and performing tasks without any noticeable issue.

I’ll come back with my findings, thank you.

1 Like