just a heads up, finally ASUS w680 available.

[The Pentaverate]

I have similar view of the speed - just wanted to confirm if anybody managed to run 4 at 4800 and not having to downgrade further. I think ASUS information about supported ECC memory is very poor - extremely few sticks reviewed (when I looked only some Innodisk that is more or less impossible to source in Europe) and no info about if even that one worked with 1, 2 or 4 sticks...
One would have imagined that they should test more generally available modules that now exists from Crusial/Micron, Kingston (this one is supposedly compatible according to vendors site) etc...

I just had a support exchange with ASUS on this, and they declined to suggest anymore DIMMs outside of the QVL. Which to me is interesting, as their QVL only lists 244 DIMMs and the marketing claims to support over 550 different DIMMs. Also, none of which appear to even be ECC-UDIMMs and there is only 1 configuration that supports 4 non-ecc UDIMMS.

@tsm37, I've been burned in the past straying away from the QVL with ASUS and would be curious of your findings if you do decide to upgrade to 128GB.

It would be interesting to know which vendor has the best maintained QVL list

 

[zettiness]

Yeah, they took like two weeks to get back to me about the RAM and then basically said the same thing, recommending those Innodisks, which I said we didn't have in Europe in my first message to start with. One of the worst customer services ever, even Intel's is better.

Also, while I was still waiting for it to answer in regards of the M.2 wireless module compatibility, the spec web page has been updated, now giving more specifics: "M.2 slot only (Key E, CNVi & PCIe)". CNVio is pretty much a dead standard with only 3-4 outdated cards available, so maybe this is CNVio2? There is no backward or forward compatibility.. At least it supports PCIe. I read that CNVio uses CPU and NIC logic to handle the traffic, so it's more efficient but what does it really translate into? Maybe for the power requirements, which would make sense in a laptop?

 

[The Pentaverate]

Even the recent w790 has over 1000+ tested and validated components but they only list 160 components on the QVL. What probably is more interesting is the claim of 1000+ validated devices and over 8000+ validation hours, which seems extremely low at 8 hours per component.

I can only assume ASUS makes these claims to inflate/justify their premium pricing.

 

[jgreco]

over 8000+ validation hours, which seems extremely low at 8 hours per component.

I can only assume ASUS makes these claims to inflate/justify their premium pricing.

You may be thinking that this means "we put these components together and ran them for 8 hours and they were fine".

I suspect it means, instead, "we fully populated this board in the most stressy configuration possible and checked the signal integrity of each memory bus trace using a scope or other specialized testgear, and found signal levels to be within tolerance".

Out here in the real world where it is impractical to do such signal integrity analysis, it is advisable to run memtest for at least hundreds of hours; here in my shop where we've been building servers for decades, we often do a thousand while burning in a system.

 

[The Pentaverate]

You may be thinking that this means "we put these components together and ran them for 8 hours and they were fine".

I suspect it means, instead, "we fully populated this board in the most stressy configuration possible and checked the signal integrity of each memory bus trace using a scope or other specialized testgear, and found signal levels to be within tolerance".

Out here in the real world where it is impractical to do such signal integrity analysis, it is advisable to run memtest for at least hundreds of hours; here in my shop where we've been building servers for decades, we often do a thousand while burning in a system.

I’m not assuming either approach, they have stated that the DIMM vendors do the testing.

Since it’s a fixed ratio it’s just very suspect.

 

[jgreco]

I’m not assuming either approach, they have stated that the DIMM vendors do the testing.

Since it’s a fixed ratio it’s just very suspect.

I think we'd need some additional clarity on this. How would ASUS know how many hours of validation a DIMM vendor has done? Why would we assume that the same validation procedure was used by each vendor?

I don't even believe that there are a thousand different DIMM vendors who would have each independently done such testing on a board, so something doesn't pass the sniff test here.

 

[Kthwaits]

I finished my build with this mobo and a 12900K this past week and everything is up and running correctly. Currently have two of the 32GB Micron (MTC20C2085S1EC48BA1) sticks installed and memtest shows that ECC is enabled. I may add an additional two sticks once they're back in stock but I'm in no rush at the moment.

The only trouble I'm having now is figuring out the right Linux drivers to detect the onboard temperature and fan sensors. I'm running UnRaid 6.11.5 which uses a Linux 5.19 distribution kernel, and when I run the 'sensors' command from the terminal I can only see the CPU via coretemp and my NVME temperatures. I can't see any of the PWM info or mobo temps (including probe I have connected to the T-Sensor header)

Is anyone running a Linux-based OS (including TrueNAS) able to detect the PWM and temp sensors from this board? If so, which kernel version are you using and what (if any) additional drivers did you have to install? I know this is a TrueNAS forum so not looking for help installing them or compiling a custom kernel for UnRaid, I can find guides for that elsewhere. I'm just trying to figure out what specific Linux drivers are required to detect this board's sensors or if such drivers even exist at all.

 

[Kthwaits]

Also, I didn't see it mentioned anywhere else in the thread but ASUS released a new bios version (2305) for this board last month.
https://www.asus.com/motherboards-c...elpdesk_bios/?model2Name=Pro-WS-W680-ACE-IPMI
Download page for W680-ACE BIOS 2305
Download page for W680-ACE IPMI BIOS 2305

Release Notes:
•Improve system performance and security.
•Support ME update function
•Improve AURA SW experience
•Improve DRAM stability

The improvements to system performance and DRAM stability are pretty vague so not sure exactly what's changed there, but I updated my board without issue. I hadn't experienced any performance/stability issues to begin with but the update didn't seem to introduce any new problems either.

 

[Kthwaits]

I finished my build with this mobo and a 12900K this past week and everything is up and running correctly. Currently have two of the 32GB Micron (MTC20C2085S1EC48BA1) sticks installed and memtest shows that ECC is enabled. I may add an additional two sticks once they're back in stock but I'm in no rush at the moment.

The only trouble I'm having now is figuring out the right Linux drivers to detect the onboard temperature and fan sensors. I'm running UnRaid 6.11.5 which uses a Linux 5.19 distribution kernel, and when I run the 'sensors' command from the terminal I can only see the CPU via coretemp and my NVME temperatures. I can't see any of the PWM info or mobo temps (including probe I have connected to the T-Sensor header)

Is anyone running a Linux-based OS (including TrueNAS) able to detect the PWM and temp sensors from this board? If so, which kernel version are you using and what (if any) additional drivers did you have to install? I know this is a TrueNAS forum so not looking for help installing them or compiling a custom kernel for UnRaid, I can find guides for that elsewhere. I'm just trying to figure out what specific Linux drivers are required to detect this board's sensors or if such drivers even exist at all.

I realize I might just be talking to myself at this point, but just FYI I was able to get sensors working using nct6775 drivers. I had to add acpi_enforce_resources=lax as a kernel flag in order for it to work though. Looks like you can fix it properly with a kernel patch but it was a lot easier for me to just add the flag and it still seems to work perfectly fine. This thread goes into a lot more detail.

 

[Etorix]

You're not talking just to yourself, but, although it has IPMI, your motherboard is more suited to a desktop workstation than to a server and few experienced TrueNAS users would be tempted to follow with hosting their NAS on W680.

 

[The Pentaverate]

Also, while I was still waiting for it to answer in regards of the M.2 wireless module compatibility, the spec web page has been updated, now giving more specifics: "M.2 slot only (Key E, CNVi & PCIe)". CNVio is pretty much a dead standard with only 3-4 outdated cards available, so maybe this is CNVio2? There is no backward or forward compatibility.. At least it supports PCIe. I read that CNVio uses CPU and NIC logic to handle the traffic, so it's more efficient but what does it really translate into? Maybe for the power requirements, which would make sense in a laptop?

The AX411 wireless adapter works, which probably means the ax211 works as well.

Also I wasn't able to find the micron DIMMs, but the sk hynix ecc-udimm (4x32GB HMCG88MEBEA081N) seem to work and pass memtest86+.

 

[jgreco]

The AX411 wireless adapter works, which probably means the ax211 works as well.

This appears to be drifting offtopic and away from TrueNAS; TrueNAS does not support wireless adapters. You are encouraged to use the Off-Topic forum for discussion of this kind of stuff. If you need any posts moved there, you may use the "Report" function to request a move of either posts or threads, just be sure to give a short summary when you do so.

 

[The Pentaverate]

VM passthrough to share Bluetooth and/or wireless…

 

[The Pentaverate]

Also I wasn't able to find the micron DIMMs, but the sk hynix ecc-udimm (4x32GB HMCG88MEBEA081N) seem to work and pass memtest86+.

Looks like i can't recommend 4 sticks of this model, Had a couple of freezes and did a multiple passes of memtest86/+ and would fail after a few. I will drop it down to two sticks and re-run, maybe it's just a bad stick, but I've never been a fan of ASUS on their poor QVL, i should have followed my gut.

 

[Ian Dawson]

You're not talking just to yourself, but, although it has IPMI, your motherboard is more suited to a desktop workstation than to a server and few experienced TrueNAS users would be tempted to follow with hosting their NAS on W680.

Do you mean any W680, or just that ASUS one? ASRockRack seem to have some interesting W680 boards for example.

 

[Whattteva]

VM passthrough to share Bluetooth and/or wireless…

I can kinda' understand the bluetooth part, but why would you use wireless for a server ever?
Presumably, you're going to the W680 chipset for the ECC support. I don't see why you'd bother going the extra mile for ECC and then cripple the platform with WiFi... You may as well just use the Z-series chipset at that point.

 

[Etorix]

Do you mean any W680, or just that ASUS one? ASRockRack seem to have some interesting W680 boards for example.

AsRockRack indeed has many interesting boards. As for W680 I mean "any W680 board, from any manufacturer", simply because it's too new. Neither CORE nor SCALE have proper scheduler support for the hybrid architecture in Alder Lake and newer.
Also anything involving DDR5 is expensive, the RAM itself, especially of ECC variety, as well as matching motherboards.

Older CPUs running DDR4 are much more attractive for a NAS, and second-hand DDR4 RDIMM is cheap.

 

[psynoz]

So I have this board as well, but with 13700K cpu. I cannot get the board to detect my storage controller(s) in the top 2 PCIe ports. The buttom 2, fine, not problem.

If I insert other cards into the 2 ports, no problem.

Did any of you have issue with detection of specific cards in those ports on this board?

PCIEx16(G5)_1	ASUS PIKE II 3008 (IT mode)	PCIe Gen 3.0 x8
PCIEx16(G5)_2

 

[Daniel15]

Asus recently announced a MicroATX version of this motherboard: Asus Pro WS W680M-ACE SE. They announced it around three weeks ago but I didn't see it until Newegg listed it for backorder last week. The differences compared to the full size ATX board are that it has:
One PCIe 5.0 x16
One PCIe 4.0 x4
One PCIe 3.0 x1
Two M.2 PCIe 4.0 x4 slots
Onboard IPMI + Gigabit Ethernet for management instead of a separate card
Two fewer USB ports (no back USB 2.0 ports)

Everything else looks the same to me.

 

[Daniel15]

Neither CORE nor SCALE have proper scheduler support for the hybrid architecture in Alder Lake and newer.

I'm not sure which Linux kernel version SCALE is using as I don't use it yet, but Intel's hybrid architecture (performance + efficiency cores) is mostly supported in the scheduler in Linux 5.18, and fully supported in Linux 6.0. Some info about the patch that landed in 5.18 is here: https://www.phoronix.com/news/Intel-HFI-Linux-v2-2021.