• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Threadripper 3000 / TRX40 and ECC?

Zarathustra[H]

Extremely [H]
Joined
Oct 29, 2000
Messages
42,211
Hey everyone,

As I am transitioning my old Threadripper 3960x and Asus ROG Zenith II Extreme Alpha board from HEDT duty to dedicated workstation duty, I am considering putting some ECC RAM in it.

But I can't seem to confirm if this is going to work or not.

AMD's webpage (and Wikipedia AMD CPU list, say nothing about ECC. Asus motherboard page claims Unbuffered ECC support (but unclear if this means that ECC RAM modules will function as RAM without ECC reporting, or if full ECC RAM capabilities will function. Motherboard manufacturers have certainly pulled this trick before...)

Googling the subject I find mixed reports of things working or not working, and most data from this era is (of course) from AM4, not from TRX40.

Anyone have any good info regarding whether or not it should work, and good ways to test functionality? Apparently confirming true ECC functioning is apparently notoriously difficult.

Appreciate any input anyone may have.
 
You can use standard DDR4 or ECC DDR4 with this platform. I ran standard DDR4 memory on my old 3970x.
 
While I think ECC RAM might be supported, I'm not sure it's actually functioning in ECC mode because it's a 3960x Threadripper (e.g. a NON-pro CPU). In other words they're physically and electrically compatible but might not be performing as intended.

I built 2 3975WX (32-core Threadripper Pro) machines for work but it's been a while. As far as confirming that ECC 'worked' on my platform as I recall the DIMMs were flagged as ECC in the BIOS after first POST, so I was pretty confident but didn't go past there.

edit#1: I think I remember the key difference between pro and non-pro for this gen - non-pros are limited to quad channel memory while pro versions use 8-channel.

edit#2: If you believe that AI is all that and a bag of chips:

Key points about ECC and the 3960X:
  • ECC UDIMM Support: The 3960X can utilize ECC UDIMMs for error correction.

  • No RDIMM Support: It's crucial to avoid using RDIMMs, as they are not compatible.

  • Potential for Enhanced Data Integrity: ECC memory helps to detect and correct bit errors, improving data reliability, especially in demanding workloads.

  • Motherboard Compatibility: The specific motherboard used with the 3960X needs to support ECC memory functionality.

  • UDIMM vs. RDIMM: UDIMMs are generally less expensive and commonly found in consumer-grade systems, while RDIMMs are designed for servers and enterprise applications.
In summary: The 3960X can use ECC UDIMM for error correction, but it's important to ensure the memory is unbuffered and compatible with the chosen motherboard.
 
To follow up, for what it is worth, I bought this 128GB OWC kit, with 4x 32GB ECC UDIMM's.

I was going to order Crucial kits, but from the reviews out there it sounds like their quality may have been slipping. They also do not offer 4-way kits that have been tested together, but OWC does. There were several reviews of Crucial DDR4 ECC UDIMM's suggesting they couldn't get all 4 sticks to work together.

This OWC kit wound up costing just a little bit more, but it may have been worth it for the "it works out of the box" peace of mind.

So, I popped the ram in, and set everything to "Auto" except the RAM speed configuration which I set to DDR4-3200, and all of the RAM was detected in POST and just booted up.

Best way I have found to confirm whether or not ECC is indeed working on an AMD systems is to - under Linux - look for whether or not the kernel has automatically loaded the amd64_edac kernel module.
Code:
$ lsmod |grep -i edac
amd64_edac             61440  0
edac_mce_amd           28672  1 amd64_edac

Looks like it has.

This can be further confirmed by looking for ECC initialization messages from the amd64_edac kernel module in the system messages:
Code:
$ dmesg |grep -i edac
[    0.403178] EDAC MC: Ver: 3.0.0
[    5.610538] EDAC MC0: Giving out device to module amd64_edac controller F17h_M30h: DEV 0000:00:18.3 (INTERRUPT)
[    5.610542] EDAC amd64: F17h_M30h detected (node 0).
[    5.610544] EDAC MC: UMC0 chip selects:
[    5.610545] EDAC amd64: MC: 0:     0MB 1:     0MB
[    5.610547] EDAC amd64: MC: 2:     0MB 3:     0MB
[    5.610549] EDAC MC: UMC1 chip selects:
[    5.610550] EDAC amd64: MC: 0:     0MB 1:     0MB
[    5.610552] EDAC amd64: MC: 2:     0MB 3:     0MB
[    5.610553] EDAC MC: UMC2 chip selects:
[    5.610555] EDAC amd64: MC: 0:     0MB 1:     0MB
[    5.610556] EDAC amd64: MC: 2: 16384MB 3: 16384MB
[    5.610558] EDAC MC: UMC3 chip selects:
[    5.610559] EDAC amd64: MC: 0:     0MB 1:     0MB
[    5.610561] EDAC amd64: MC: 2: 16384MB 3: 16384MB
[    5.610563] EDAC MC: UMC4 chip selects:
[    5.610564] EDAC amd64: MC: 0:     0MB 1:     0MB
[    5.610565] EDAC amd64: MC: 2: 16384MB 3: 16384MB
[    5.610567] EDAC MC: UMC5 chip selects:
[    5.610568] EDAC amd64: MC: 0:     0MB 1:     0MB
[    5.610570] EDAC amd64: MC: 2: 16384MB 3: 16384MB
[    5.610571] EDAC MC: UMC6 chip selects:
[    5.610572] EDAC amd64: MC: 0:     0MB 1:     0MB
[    5.610574] EDAC amd64: MC: 2:     0MB 3:     0MB
[    5.610575] EDAC MC: UMC7 chip selects:
[    5.610577] EDAC amd64: MC: 0:     0MB 1:     0MB
[    5.610578] EDAC amd64: MC: 2:     0MB 3:     0MB

Success there too.

I'd say I have 128GB fully functioning ECC!

Before calling this a win, I plan on running a memtest overnight, but thus far these are good signs!

Thanks for the help guys!

I'll probably post back once I have confirmed with a memtest run that everything is as it should, but that may take a little bit, as I recall from the last time I did it the standard test is four passes, and as I recall each pass takes quite a while. On my server each pass took several hours, but to be fair that was with 512GB of Registered ECC on a lower clocked server CPU.

Anyway, I'll start it before bed, and I guess I'll find out how long it took tomorrow morning :p
 
As an Amazon Associate, HardForum may earn from qualifying purchases.
Back
Top