Sorry, misread your post entirely initially. Yes, no difference as far as I'm aware between regular desktop memory and "non-ECC UDIMMs" cos that's exactly what desktop memory almost always is. Assume this is for workstation-esque boards that can take things like the i9 series?
Well depends on your definition of "server" Personally I only ever use ECC in boards that are going into use as a server, since any data that might be sitting in a memory cache somewhere might possibly get a bit flipped, so non-ECC is less than ideal from a data integrity point of view.
Last I checked ECC had only a slight price premium over regular non-ECC stuff (i.e. ECC and non-ECC both cost an arm and a leg but buying non-ECC might get you some toenails as change). Most of the extra cost typically comes from having to buy a motherboard and CPU that are compatible, so if you're already buying an ECC-capable chip and mobo I'd seriously consider going the ECC memory route (and I suspect a goodly proportion of the members here would echo that as we deal primarily in servers where price generally takes a back seat to reliability). However all of this depends on your use case, overall budget and what market segmentation Intel has come up with this week of course
non-ECC UDIMM is exactly "normal" desktop memory.
The xeons which use the same socket as the non-HEDT platform (LGA 115x), along with a supporting motherboard, will take ECC UDIMM.
The HEDT platform xeons (LGA20xx) will take RDIMMs or ECC UDIMMs or non-ECC UDIMMs. The most memory-versatile systems.
RDIMMs are always ECC and include a buffering chip, this is what you use on the "high end" server platforms and their spawns.
memory is really a pain in the ass to shop for on some sites because of this separation.
Personally, I doubt I'll ever buy/build another "desktop" type machine that doesn't have ECC. As others have said, the price premium of the memory itself is negligible, and the platforms are typically more reliable anyway. The main cost differential is in the CPU and motherboard, and with a surplus of used gear floating around that is plenty fast enough for me (E5 v2), that becomes a nonissue as well.
When Windows BSODs without ECC, how do you know it wasn't a memory error? Which module did it affect? I know ECC doesn't make a machine immune to BSODs, but it can assist in troubleshooting.
Hi all, thanks for the feedback. After having Supermicro X9SCM-F, it's my first taste of server hardware, with IPMI. I can definitely say I'm not going baaaack But reason I'm asking is, considering new Supermicro serverboards. Some Xeon D boards offer ECC/ non ECC UDIMM. 2nd hand DDR4 ECC memory aren't that plenty. To buy new are still quite expensive. So since I'm gonna use it for home NAS, I figure non ECC UDIMM is alright, hence my question.