I was planning to build a Zen 5 PC with ECC memory and I was wondering if I would have any problems with ASUS ProArt B650-CREATOR and two Micron DDR5-5600 ECC UDIMM sticks, since they’re not on the QVL list.
I don’t know if waiting for a B850 board would be worth it. I don’t know what I would gain, aside from maybe not having to update the BIOS. Also, for how much longer do you think is the B650 version going to be produced, once B850 boards are released?
Should be fine in principle. In practice, I’d plan for validating SECDED with every BIOS version the board runs. Don’t know of any ProArt B650 ECC tests but Wendell recently found the ProArt X870E wouldn’t post with ECC DIMMs without a BIOS update. At least one ROG X870E’s failed to deliver claimed ECC support as well.
ASRock’s ECC delivery seems more consistent. The boards are a good bit more expensive but, if you need 5 or 10 GbE down, Gigabyte’s ECC support on their upper end B850 and a couple of the upper end X870Es might also be worth considering.
Wouldn’t surprise me if B650 production stopped last year.
You mean testing it with something like memtest86 (I don’t know if it supports ECC on Zen 5 yet)? Or just checking if it’s reported as ECC by the kernel? Is there a safe way to check if there are correctable errors reported other than overclocking it?
Going with ASRock’s Taichi Lite was exactly my plan for a while. It was even one of the few boards which still had 6 SATA ports, which I was also looking for. However, I recently learned that ASRock boards unfortunately don’t offer individual SATA port enabling/disabling like ASUS does which is quite important for me. I then saw people reporting here and on Reddit that ECC worked on ProArt, mostly X670E but I think I saw two users which also said it worked on B650 (actually, one of them was B650 TUF, I just found it here).
Really? I thought stopping their production would be announced somewhere online and I couldn’t find it. Because I don’t know if a B850 board will be available for a similar price soon.
Masking pins is probably the only reliable way unless low cost DDR5 fault injector risers become available. Multiple folks here have tried overclocking and not gotten errors. I think also two folks have tried the fault injection in MemTest86’s paid version somewhat recently and been unable to confirm it’s testing anything.
With ECC reports from Level1 forum folks people usually are at least checking the logs but it’s unsurprisingly routine that there are no errors and thus no confirmation SECDED’s working. Seems like elsewhere reports of support may be based on as little as booting, without checking ECC’s enabled in the BIOS or that the OS reports it’s at least seeing the ECC and monitoring for SECDED events.
With Epyc 4004 AMD seems to be in a position where breaking ECC in an AGESA release, as they did shortly after Zen 4 launch, is pretty unlikely. However, I’ve found it really hard to locate data on the extent to which board partners treat ECC support on AM5 as a high reliability feature versus a marketing checkbox. Anecdotally, Asus seems to have the highest incidence of oops, nope, board speced for ECC doesn’t actually feel like ECCing at the moment.
As I’ve commented in various other ECC threads, I find it hard to articulate a value proposition to adding bus EC4 over non-ECC DDR5’s write CRC + on die EC2 + read CRC. Partly that’s the difficulty of confirming EC4’s consistently SECDEDing across BIOS versions, partly it’s that I do a lot of test passes on non-ECC UDIMMs that find no errors.
Interesting. What’s the use case?
For a January 15th globally coordinated launch the boards have to be produced in time for container packing, container loading, shipping, unloading, unpacking, warehousing, and retail distribution. For southeast Asia to Europe that’s like at least a six week lead. So production has to start shifting around mid-November.
But I’ve never seen anything on how mobo manufacturers choose to allocate production between old and new models. From the outside, so without any sales data, it’s also hard to differentiate between ongoing production of popular models and old stock that’s not moving in retail. I’d be surprised if there’s not also significant region to region variation.
FWIW, from what I can tell from how ASRock availability here’s gone the past few years pretty much all X boards and many of the Bs switch out fairly quickly. Seems like a few of the B models stay in limited production, more in mATX than ATX, which is consistent with how I’d guess sales volumes evolve over time. I suspect probably there’s occasionally still runs of X570 Taichis.
I seldom build Asus boards so don’t have a sense of how their sustaining availability works.
Me neither but I’m not counting on it. A business objective of the X870 and X870E refresh cycle seems to be to reset pricing by treating it as a new launch rather than a refresh. What’s been posted of pricing here suggests the same for B850, though the MSRP increases are lower.
I restrict which system can access which drives, since I multi-boot. I still haven’t found an alternative to this method which I’m using right now. I was quite surprised when I learned this isn’t a standard feature on most boards. From what I saw in BIOS manuals it’s mainly available in ASUS models (and Gigabyte I think).
I’m not entirely sure if I understood what you meant here. If you were referring to whether ECC DDR5 is worth it over non-ECC, the cost difference is primarily in the memory sticks which I’m fine with so I want to make sure the motherboard will work properly as well. I know these errors shouldn’t happen frequently, but since I process a lot of data I still want to do as much as I can to keep the risk of undetected silent corruption as low as possible.
Really? I just remember ASRock removing ECC from their boards’ specification (and the BIOS options maybe, not sure) about a year or two ago. But it seems like it was added back later. I thought such a thing isn’t usually broken in an update. Besides, as long as I update to a version which supports the CPU and ECC and stick with it I should be good right?
This seems to be supported on my ASRock B650M-HDV/M.2 at least, although unintuitively you need to go to the chipset (Promontory) settings rather than the Storage settings:
I also dual boot and want to prevent my gaming OS from accessing my “important” drive. I use a kernel option to disable the SATA port in question (libata.force=1.00:disable). I’m sure it can be circumvented but at least the drive isn’t immediately accessible by the OS without further effort.
Sjould be fine, do note that CREATOR board uses Realtek NICs which may not be the best choice depending on application. While being a bit more expensive you might want to have a look Asus ROG Strix B650E-E or ROG Strix B650E-F depending on your requirements.
Interesting, I thought ASRock boards in general didn’t have that option (by the way, did you ever actually use that option and see that the SATA port becomes unavailable?). However, I remember reading somewhere that whether SATA ports can be disabled individually depends on whether they’re connected to the AMD chipset directly or to an ASMedia controller. That would explain how your board has the option for the 4 SATA ports connected to the chipset. But unfortunately I don’t think Taichi allows you to disable even the four ports connected to the chipset (the other four use ASMedia). I couldn’t find in the manual and I read some people writing about it (here I think).
Actually, I just looked in both X670E and X870E Taichi’s manual and they both have the PROM21 Chipset SATA Configuration Options as well, but it says nothing more about SATA port configuration. It might actually have the same option as your board with no mention of it in the manual? I’ll have to ask people with these boards to check.
Yeah, but I want to do it on a lower level. And I boot Windows as well and I don’t think it gives you such options (or that I would trust them if it did). I’ve been doing it in BIOS so far and it’s worked perfectly for me.
I tried it now. The port is still detected, but the kernel reports SATA link down. So it went from this
[ 1.599634] ata3: SATA max UDMA/133 abar m1024@0xf6980000 port 0xf6980200 irq 46 lpm-pol 0
...
[ 2.571242] ata3: SATA link up 6.0 Gbps (SStatus 133 SControl 300)
[ 2.574364] ata3.00: supports DRM functions and may not be fully accessible
[ 2.575339] ata3.00: ATA-11: Samsung SSD 860 EVO 500GB, [redacted], max UDMA/133
[ 2.576393] ata3.00: 976773168 sectors, multi 1: LBA48 NCQ (depth 32), AA
[ 2.580696] ata3.00: Features: Trust Dev-Sleep NCQ-sndrcv
[ 2.581632] ata3.00: supports DRM functions and may not be fully accessible
[ 2.586609] ata3.00: configured for UDMA/133
[ 2.588934] ata3.00: Enabling discard_zeroes_data
[ 2.595886] ata3.00: Enabling discard_zeroes_data
to this:
[ 1.429740] ata3: SATA max UDMA/133 abar m1024@0xf6980000 port 0xf6980200 irq 46 lpm-pol 0
...
[ 2.403340] ata3: SATA link down (SStatus 0 SControl 300)
(Now I just need to remember to enable that port again at next reboot. )
Yeah, that’s what I meant. That’s what happens on my current ASUS board. Hopefully everything works the same way. I’ll have to check somewhere if Taichi has this option as well. I might as well go with ASRock in the end, but it will be double the price of this ASUS B650 board so if ECC works properly with both of them…
They’re in general more buggy and have much worse driver support than Intel however in your use case you might not notice a difference but I doubt you’ll see anyone tell you to go Realtek over Intel.
You mean disabling the ASMedia SATA ports in BIOS? I think I read somewhere that they could be disabled in BIOS as well on ASRock but either all together or none and not individually like chipset SATA. I might be wrong.
Now I’m really not sure what I should do. I’d like to go with ASRock, but then I need to go for Taichi Lite because cheaper models don’t seem to have chipset connected SATA and it looks like only those can be individually disabled. But it is quite a bit more expensive and I don’t need most of its other (price increasing) features.
If I go with ProArt, I lose two SATA ports but it’s less than half the price of Taichi. I get PCIe 4.0 instead of 5.0, but PCIe 5.0 on Taichi would drop to x8 anyway when I add a second PCIe card. On ProArt I can add it to the chipset’s PCIe slot without affecting the GPU’s slot. I also get a second Ethernet port which I won’t use most of the time, but it may come in handy every now and then. And most importantly, ECC should hopefully work on the updated BIOS?
Seems like a tough call since these are the only two options that I can see right now…
Yes. An ASM1061 gives a pair of ports so, if disablable, it’s not that much of a constraint on how legacy drives are laid out. Most of the enable-disable switching would presumably be between NVMes.
Most X870Es do x8/x8 bifurcation, quite a few with one or two chipset slots as well, and dual NICs aren’t that uncommon. All mobos have tradeoffs and, if there isn’t a clear decision among them, usually that’s an indication the exact choice doesn’t matter that much.
Hopefully. But if hopefully’s good enough then probably non-ECC’s fine too. Some folks find peace of mind in that niche but it’s usually difficult to articulate much of an engineering value proposition for it.
I hope the memory I’m planning to get would work on Taichi. I looked in its QVL and it does list the Micron ECC UDIMMs, but only the 4800 version of the same model. I hope 5600 works as well.
I don’t understand what you mean by this. Of course I use different filesystems. But that doesn’t prevent an OS from accessing all the connected drives.
NEMIX appears to be sourcing from random places and I’ve found several posts on different forums mentioning that SPD is and other memory date is simply blanked, missing or DOA so I’d rather stick with the major brands such as Crucial/Micron/Samsung/Kingston etc which has yet to fail on as long as JEDEC specs are followed despite not being on any QVL list.