As I don’t want to necrobump an old thread, here is a new one.
My question is pretty much asked in this thread, though it isn’t answered to the full extent that I need it to be answered.
The new PC will be based on the
X570S Aorus Master Ryzen 5800X or 5900X WD SN850 1TB with my old RX480 GPU for now.
Now the big question: As I’ve heard, the best setup would be two DIMMs dual-rank and clocked as high as possible. By my personal request it should be ECC. My hope is to get more stability during frequent suspend mode use. Also I’ve always been curious about ECC in my PC. Which is now possible with Ryzen.
What modules should I use, as the whole system shall have 32GB, which means 16GB sticks, supposedly 3200Mhz and dual-rank. This boils it down to two options:
Kingston Server Premier DIMM 16GB, DDR4-3200, CL22-22-22, ECC (KSM32ED8/16HD) with Hynix D-die
**and **
Samsung DIMM 16GB, DDR4-3200, CL22-22-22, ECC (M391A2K43DB1-CWE) with unknown modules, but I’m quite sure those aren’t Samsung B-die
Micron E-die is available from Kingston, but only as a 32GB module and the Samsung B-die chips aren’t available anymore AFAIK.
So which one to choose? Or are there other alternatives that are better suited for the task?
Anyone any experience with these modules and Ryzen 5000?
I use the Samsung ones, 4 sticks of them, and all have chips with the marking “5WD BCWE” from my research those are Samsung B-die chips
I don’t know if this is any relevant but mine have an extra letter at the end: “Q” so their full name is: “M391A2K43DB1-CWEQ”
As of now they work in an 3960X on an ASUS Zenith II Extreme albeit it seems the board always needs a second attempt when cold booting. You can read what I’m experiencing here, if you want.
Afaik Kingston 3200 16GB are Hynix D-Die, but I’m not fully sure about that.
If you have no intention of overclocking any of these modules and want to run them all on stock it kind of doesn’t matter which manufacturer you go with. All will run on JEDEC default.
Kingston has a very good website where you can enter your motherboard/system and it spits out compatible memory. The server premier line entries are all listed with dram assembly both on the website as well as in pdf data sheets.
I did check with Kingston for TRX40 boards some time ago.
edit: there seems to be a 16GB variant (SM32ES8/16ME) with Micron-E, but is listed as being replaced by some other memory that has a dead link when clicked on
Get 32GB sticks There is no such thing as too much memory.
In the end I bought the Kingston Hynix D-die 2R-8 which run now according to JEDEC. They were automatically detected and run with the standard timings since.
Which seems to work fine, but I’d like to confirm that they are working / really correcting errors. Inxi, memtest and the BIOS report them as such, but do they really?
I found a forum post on how to do so a while ago, but I don’t seem to be able to find it back again. Could one of you post the Linux CLI commands to do so? I would be very grateful.
The 32GB sticks were prohibitively expensive, unfortunately…
From what I agethered about ECC correcting erros it seems that the motherboard needs to have something enabled to report it to the OS, not all seem to do it.
