On this board, it appears
BLUE = BANK 0
WHITE = BANK 1
percent
Agreed matching pairs is best practise.
I discovered during testing that one of the modules was made in Dec 2018 while the other 3 were made in Mar 2016ā¦the younger PCB looks slightly different in colour, but technically it is the same specification (AFAICT), so it shouldnāt matter in theory. It made me wonder if they were returned or whether they just sat around on the warehouse shelf for agesā¦there is no way to know
Re: point 3) I tested a single module in all four slots without error, but MemTest86 testing may not be the same as booting up in TrueNAS (or Windows etc.) - but the indication to me is that this is not necessarily true. I assume BIOS interrogates all slots during POST (how else would it know all four slots are in use?).
I let it run for a good while - as I said, when it errors, it errors within seconds. No errors during Tests 1 to half way through Test 9. If it was gonna error, I would have seen errors well before this.
Currently testing with two sticks, one in A1 (BLUE) and one in A2 (WHITE). This is not in the list of recommended combinations in the manual, but again, nothing. Nada. Zilch. Zero errors so far.
This is why I really believe it has to do with voltage and memory timings (CAS, tRCD, tRP, tRAS, etc.)ā¦or (perish the thought) some bad soldering on the boardā¦but I donāt think that is the case, else I would have expected errors to always occur for a given slot regardless of whether I had 1, 2 or 4 modules inserted.
No such setting - just āECC Enabled/Disabledā.
I donāt want to run it with ECC disabled, since that was the whole point of choosing this board and these sticks.
I think I have one bad DIMM slot that only errors when 2 or 3 other slots are filled, but does not error when only 0 or 1 other slots are filled.
I think Iām gonna call it quits and settle for 3 sticks and get on with my life.
I canāt RMA the board or even buy a replacement since they are no longer available.
If anyone knows of a current ITX sized board that has 13 SATA ports and supports 64GB or more of ECC DDR4-2400 memory, please let me knowā¦
If you can drop that to 12 SATA ports (with no further adver consequences on other shared ports being disabled), Supermicro A2SDi-H boards qualify. You can certainly order one through an official Supermicro retailer down-under, since these boards are still current. As for finding one available, in Australia and for a reasonable price, wellā¦
I originally built it using 12, so thatās not out of the question.
I went with ASRockRack over Supermicro because I have had an, possibly unfounded and inconsistent, aversion to Supermicro due to early reports of CCP interference in the manufacturing and the supply chain (I even asked about this on the old forums a couple of years ago - as @naskit). I recently saw that they are now advertising āMade in USAā products. Perhaps I need to get over that and just have a GeoIP database and smart DNS to block traffic going to certain countries from inside my network.
One other thing to check: Is the RAM on the qualified parts list? If so, does it match down to the last digit on the SKU?
I have a SM board here (see signature) that would only recognize exact matches, otherwise, it would drop sticks in slots 2&3, no matter what combination of sticks I put in slots 0-3. The difference between the SKU I received and the SKU in the SM-approved list was the revision of the RAM dies used on it. I.e. the vendor, speed, etc were all correct, just the SM board wanted a older die revision than the one I received from my vendors.
Thankfully, all my vendors took their RAM back, searched their inventory, and sent me a 100% matching SKU, at which points all DIMM slots went live.
percent
