Yeah I’ve been hearing a 50-100% price hike for DRAM and NAND this year from all the usual analyst sites.
1 Like
Sad times, shipping issues probably aren’t helping
That’s 4800mhz tho. I was excited to finally have fast ECC! Ryzen benefits a lot from higher memory clock, so very annoying we can’t get any fast ECC.
How true is that these days? Like is it really that much of a gain compared to what Intel gets, I dont think I have seen a lot of emphasis on that.
Very true, if you’re talking 3600mhz vs 4800mhz+. Anyway, I found that ECC 5600mhz RAM! I’ll report back when I’m able to test it…
1 Like
I have been eyeing this DDR5 ECC UDIMM RAM from Apacer at 5600M/Ts in 32GB / stick form for my X670E / Ryzen 7950X desktop. I have an ASUS Pro Art X670E Wifi and ASUS and AMD claim that the board and processor both support ECC UDIMMs. I would try to get it to run at 6000mhz with tighter timings also. The problem is trying to find someone who will sell it to me. I tried Mouser and they were less than helpful.
1 Like
Every time I re-google “48gb ECC udimm” this thread pop ups. And I hope someone replys it with a link onto some affordable and normal priced hynix 48gb ECC udimms.
2 Likes
Mainstream availability is slowly appearing. In central Europe 2 listings for kingston modules appeared:
I have been on the lookout for almost a year, but caved in 6 months ago and got some GSkill modules instead, when RAM prices started creeping up again.
2 Likes
While it’s nice that finally DDR5 ECC UDIMMs larger than 32 GB hit the market I’m sad now that it seems that we’ll have to wait again until we eventually get the larger 64 GB ECC UDIMMs 
Thought that maybe ECC UDIMMs from the big players would jump directly from 32 GB to 64 GB
Balls.
2 Likes
Interesting has anyone tried 2 or 4 dimms of that Kingston 48GB KSM56E46BD8KM-48HM and also how much does it overclock as does it run 3000+MHz in 2 or 4 dimm conf. Seems the best choice right now either way.
Old setup is running fine with 4 x 16GB ECC dimm M391A2K43BB1-CTD 2666-CL19 @ 3466-CL18 1.3V OC but looks like DDR5 OC capabilities might be much worse, especiallt if 4 dimms installed.
4 dimms of ddr5 you are lucky to get 4800 tbh.
I have 2 dimms of 5200 ecc (80 bit) and I am so happy its as fast and stable as it is
3 Likes
Also interested. Currently running 4x32GB ECC UDIMMs at stock frequency.
Considering downgrading to 2x48GB for better memory performance.
Hoping Ryzen 9000 comes with a higher performing memory controller, but, from what have been announced so far, chances of that are slim to none.
Said that, even at stock frequencies ram already runs on the warm side at full load. Some additional airflow and two DIMMs instead of four would help with that, me thinks.
FWIW even messing with the B650 Asrock the ECC support is still bugged on “official” bioses even 1.0.0.7c (which that version is ECC okay on some desktop boards). Had to go to 10.18 beta bios to get properly working ECC.
Do you mean that even Asrock B6xx boards with latest AGESA 1.1.7.0 have still bugged ECC and need some newer/older beta or luck with 1.0.0.7b beta version?
On many noard 1.0.0.7c 9r newer was goid enough for ecc but the akctual ipmi asrock boards need a beta bios which is newer
Since you currently have a relationship with ASRock, can you determine if their ECC shenanigans (working in one BIOS version, not working in another) are purely accidental due to mistakes from their BIOS engineers or if it’s intentional to “phase out” ECC functionality as a whole in the long-term?
As mentioned somewhere before I am highly critical of the addtional AM5 Epyc SKUs (since they are the same as the Ryzens only with some additional firmware features enabled and a different/lower boost behavior for increased efficiency), I smell Intel’s product segmentation policies from the 2010s behind that, and can see AMD possibly cutting away ECC from regular Ryzen CPUs in the process.
1 Like
right now it appears the f’up was entirely on amd’s side? The new agesa seems to actually universally enable ecc across desktop and epyc am5 cpus, finally, still leaving it up to board vendors to quality… I just don’t think anyone had really tested ecc over the last year and no one took my reports of “this is weird, even when I patch the mainline kernel to enable edac, its really weird”
consider that asrocks “server” board with agesa 1.0.0.7c did NOT enable ecc udimms whereas all the 2024 bioses for the steel legend and pro RS, not even the taichi, enabled it. still the case the b650 w/broadcom 10g board is broken except for the “beta” 10.xx bioses. bios 4.09 is f’d
new agesa asus pro art is a DREAM also. and ecc support seems good in rog strix/hero boards now too. owing to 2024 bioses. doubt any of these board vendors were really testing.
contrast to my 2023 computex vvideo asking every vendor “yo where the am5 server-ish boards at?” and being treated like a 3-eyed alien whereas in 2024 there were multiple am5 skus from gigabyte, msi, mita, wiston, asus/server, asrock and many others. including dense/blade configs.
3 Likes
Have been using ASUS AM4 motherboards since 2019 and one of their AM5 models since earlier this year and here BIOS updates never broke functioning ECC so AMD’s AGESA doesn’t seem to be the actual source of the fluctuating ECC functionality on ASRock’s end.
I’m highly hesitiant touching anything by ASRock Rack ever again, behind the scenes their firmware update processes are horrible, rivaling Broadcom’s. It would be one thing if their motherboards would be released in a “perfect state” and not regularly with many obvious bugs.
After everything I’ve seen for home lab use I’d always go with one of the most popular (not obscure) premium consumer motherboard models, currently only ASUS remains (the only guys remaining next to ASRock not bean-counting away the additional ECC memory slot traces in their motherboards). Of course It sucks here that standalone IPMI KVM AICs aren’t more mature to be an option without needing any second thought, completely independent of the specific motherboard model.
what was the am5 you had? I had the strix hero and while it booted with ecc, it never reported an error with ecc even when I did the edac kernel patches. its mainline now…
whats your output of dmesg and dmidecode -t memory on that board? (and bios versions).
I agree, something extra f’d was going on with ipmi boards because 1.0.0.7c was the first agesa I saw that had Actual Working ecc on some boards, including asus pro art am5 x670 (woo 10gb, vs b550 w/o) but … its been weird to say the least
1 Like
My Ryzen ASUS model history is:
-
AM4: Pro WS X570-ACE
-
AM4: ProArt X570-CREATOR WIFI
-
AM4: ProArt B550-CREATOR (that model has build quality issues, noticed loose USB-C ports on multiple units purchased a year apart, no issues on the very same USB-C ports of numerous ProArt X570 units)
-
AM5: ProArt X670E-CREATOR WIFI
Was late to the AM5 party and honestly think the ProArt X670E is the only “great” motherboard model out there so far, every other one is like “meh” from a PCIe or peripherals connectivity standpoint.
I still don’t understand why there isn’t a single AM5 motherboard that uses its 8 M.2 CPU PCIe lanes to instead provide an additional x8 CPU PCIe slot, in theory making a x8 (CPU), x8 (CPU), x8 (CPU), x8 (X670E) PCIe slot configuration possible.
How I check for ECC:
-
Check Windows terminal to see if it detects Ryzen’s Multi-bit ECC
-
Undervolt and overclock the memory to the edge of a system not POSTing to force actually occuring memory errors under heavy load
-
Boot up PassMark Memtest Pro, an example:
1 Like