12gen N-series Nas motherboard (topton, cwwk, ... )

[Dr.FrankenHouse]

In case you mean suggestions for a new PSU: I went for buying a 2nd hand but hardly ever used Corsair SF600 Platinum (CP-9020181-EU) which has a power dissipation of just 7 - 8W with loads of 20 - 40W. An SF450 Platinum would have been better (drawing even 1W less)[1]. From my research on the web, the Corsair SFXXX Platinum PSUs still belong to the most efficient SFX PSUs for low power builds available today.

If your Homelab case allows for a standard ATX PSU, the Corsair RM550x 80+ Gold 2021 (CP-9020177-EU) would even be more efficient, having a power dissipation of only 5 - 6.5W with loads of 20 - 40W. [2]

Unfortunately, the lower wattage Corsair SF/RMx PSUs seem to be out of production and are not available any longer (at least here in Europe/Germany). That's why I decided buying a "good as new" 2nd hand one.

[1] Corsair SF750, SF600, SF450 Platinum im Test - SFX jetzt mit Platinum-Effizienz
[2] Corsair RMX 2021 im Test - Die neue Generation der Mittelklasse!

What about PicoPSU?
My build will be (I'm waiting for some parts)
Jonsbo N2
i3-N305
32GB Crucial DDR5 5600MHz
3xWD Red Plus 8TB (5400rpm)
2xSSD Crucial MX500 1TB
2xnvme (tbd)

If my calculations are right, a PicoPSU 160w should be fine. Or not?
In alternative, sf450 or sf600?

 

[Zer0_C00L]

Rules of thumb
Hard Drive
5 w - steady state
10w - spin up (more like 7)
1 w - idle

SSD
5 w - constant

Guessing these boards use about 30w max (Def less than 40).

150w would be PLENTY.

 

[FunkStar]

Once I got the boot drive converted to GPT and configured for UEFI, it booted right up. With 1 NIC active, 1 SSD, three HDDs, and a USB Coral TPU, it maxes at about 43W and idles down to around 20 once everything spins down. Cold boot doesn't seem to go above 70W. I haven't done much in the BIOS save for enabling C-states, and powertop shows that I'm in C8/C10 80% of the time (although the report shows Pkg(OS), not Pkg(HW) so I'm not sure if these are accurate numbers).

@whoiswes How are you getting higher then C3 states? The cwwk mobo's uses the JMB585 SATA controller that does not support ASPM. So the CPU won’t go into a deep sleep.

 

[whoiswes]

@whoiswes How are you getting higher then C3 states? The cwwk mobo's uses the JMB585 SATA controller that does not support ASPM. So the CPU won’t go into a deep sleep.

Honestly, I dunno. I'm going off of what powertop is showing. Note that I'm getting these numbers with or without C-states enabled in the BIOS, so these numbers might not be valid, in which case I'd love for some input/help in figuring things out. I read the same thing about the JMB585 and was hoping someone would try some alternate controllers (looking at the M.2 ASM1166 breakouts).



FYI, I see similar stats with cpupower monitor.

EDIT - this is all fairly new territory for me so if I've made an obvious mistake, please (politely, if possible) point it out.

 

[FunkStar]

I'm kinda new with this stuff so I don't have the expertise to tell you why you're getting higher C-states then expected. But I'm very curious!
I waited to long so now the N305 isn't in stock any more (or overpriced) so I don't have mine yet, I'm just slowly trying to collect the stuff.
Maybe start listing what hardware you're using?

Also looking at this CWWK mini pc with an 4*M.2 NVME interface for placing at my parents place so I could have my offsite back-up.
But it should be all nvme storage build and no how that would work.

 

[Steffe]

Honestly, I dunno. I'm going off of what powertop is showing. Note that I'm getting these numbers with or without C-states enabled in the BIOS, so these numbers might not be valid, in which case I'd love for some input/help in figuring things out. I read the same thing about the JMB585 and was hoping someone would try some alternate controllers (looking at the M.2 ASM1166 breakouts).

View attachment 33712

FYI, I see similar stats with cpupower monitor.

EDIT - this is all fairly new territory for me so if I've made an obvious mistake, please (politely, if possible) point it out.

Regarding c states. These are the ocenes available. Package C-States - 001 - ID:655258 | 12th Generation Intel® Core™ Processors Datasheet, Volume 1 of 2
So the majority if the time you're in the lowest? If I understand it correctly...?

Well, it looks correct..
Can you share the output of

cat /sys/devices/system/cpu/cpu*/cpufreq/scaling_driver
cat /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor

Well, a system at C10 would be completely useless. C10 implies the processor is completely shutdown according to ref https://www.reddit.com/r/homelab/comments/zdbvz2 . I will using mine for proxmox and firewall, so I guess some cores always be active or below c5, whichever it is here, I assume.

What about PicoPSU?

Yes. Plenty, be sure to have enough ports. Remember that hdd has a high watt spike when they spin up. I'm getting this one. Closest wattage wise and very cheap SMART Series-Chieftec. About $50.
It will take some time to go break even for more expensive. The most efficient is this one HDPLEX 250W GaN Passive AIO ATX Power Supply. See comparison here https://www.reddit.com/r/sffpc/comments/15kqiut
RM550x is completely unavailble.

 

[whoiswes]

cat /sys/devices/system/cpu/cpu*/cpufreq/scaling_driver

All 8 cores are intel_pstate.

cat /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor

All 8 cores are powersave.

I don't know if I'm actually reaching C10 but I doubt it, based on calling cpupower frequency-info. Note the line indicating that the call to hardware failed. I'm guessing that's the problem but I haven't been able to figure out a fix.

EDIT - mix freq is 800, so no way I'm hitting any of the deeper C states. No idea why powertop and cpupower report differently.

analyzing CPU 0:
  driver: intel_pstate
  CPUs which run at the same hardware frequency: 0
  CPUs which need to have their frequency coordinated by software: 0
  maximum transition latency:  Cannot determine or is not supported.
  hardware limits: 800 MHz - 3.80 GHz
  available cpufreq governors: performance powersave
  current policy: frequency should be within 800 MHz and 800 MHz.
                  The governor "powersave" may decide which speed to use
                  within this range.
  current CPU frequency: Unable to call hardware
  current CPU frequency: 800 MHz (asserted by call to kernel)
  boost state support:
    Supported: yes
    Active: yes

 

[youddha]

Anyone found a solution for nvme radiators/heatsink ? Not sure how to fit 2 of them side by side...

 

[chripo]

Thanks! Yea, my case ventilation should be very good. And yep, I would definitely have a heatsink on the copper plate . I was thinking of just getting the bundle with the Jonsbo radiator and try it with and without the fan (but with the heatsink still on it). I wasn't able to find any cheap LGA115x heatsinks without fans, so for the $17 extra (to get the Motherboard + Jonsbo bundle), I figured I might try it with the Jonsbo and just unclip the fan. Worst case scenario is I put the fan back on, and if it's too noisy, I can always swap out the fan.

I don't know how much the Jonsbo cooler bundle is on top on CWWK/AliExpress stores currently, but from what I have researched on the net the Noctua NH-L9 will definitely perform better. I configured the fan in the BIOS to be something like semi-passive, so it only starts up with very low speeds on higher CPU loads/temperatures. I can't hear it even when HDDs are off, but alas, I'm an old guy. Anyway, I bet it would be no problem to run my system even without the CPU fan. But please consider that I re-pasted the original copper plate using Liquid Metal paste. Unfortunately, I didn't measure temps before that. For the Noctua cooler I just used the paste they included with it.

What about PicoPSU?

150w would be PLENTY.

Yes. Plenty, be sure to have enough ports. Remember that hdd has a high watt spike when they spin up. I'm getting this one. Closest wattage wise and very cheap SMART Series-Chieftec. About $50.
It will take some time to go break even for more expensive. The most efficient is this one HDPLEX 250W GaN Passive AIO ATX Power Supply. See comparison here https://www.reddit.com/r/sffpc/comments/15kqiut
RM550x is completely unavailble.

Completely agree, 150W is plenty. My smart plug never measured anything higher than 60W with 4 HDDs and an SSD, and I did run stress tests like Prime95 with Small FFT on the system. But there might be spikes on e.g. power-on which my smart plug does not record.

I was also thinking about a PicoPSU when planning my build and depending on your use case, this might definitely be the best choice. What I personally don't like about PicoPSUs is that they need an external power brick and are quite limited regarding connections. In my case, with the Jonsbo N3, I needed 2 x MOLEX, 1x SATA, 1 x 24 pin CPU, 1 x 4 pin EPS/12V CPU. It might be possible to accomplish that with a PicoPSU, some adapters and a good power brick. But personally, I didn't feel well with such a setup, especially because the Jonsbo N3 already features a mounting space for an SFX PSU and would have been modded (honestly no big deal) for using a PicoPSU with a DC power brick.

The HDPLEX (and possibly other GaN PSUs) compared to PicoPSUs might have been a better option, but I also couldn't find a nearby dealer easily. So finally, I went for the easy solution and expected ~3-4W additional power dissipation with my 2nd hand SF600 Platinum. BTW, at least from my price research here in EU/Germany PicoPSU + power brick or HDPLEX GaN are not that much (if at all) more cost efficient compared to a Corsair SF450/600 or RM550x, even on a longer (~10 year) timeline. But yes, you have to be lucky to get one of these 2nd hand in good/nearly unused condition today.

For those who are also unsure about PicoPSUs/GaN-PSUs and have an ATX case: I would like to second @youddha 's recommendation, the be quiet Pure Power 11 FM 550W. This should still be available at stores and would have been my 2nd choice after the RM550x for ATX.

 

[audelair]

Anyone found a solution for nvme radiators/heatsink ? Not sure how to fit 2 of them side by side...

I think that those M.2 slots are only Gen3x1 speeds… so you’d be getting a Gen3 drive and shouldn’t worry about adding any heat sinks, IMO.

 

[youddha]

I think that those M.2 slots are only Gen3x1 speeds… so you’d be getting a Gen3 drive and shouldn’t worry about adding any heat sinks, IMO.

Yeah but I'm getting temperature above 40°c on one of the slot quite consistently. Which trigger the default warnings in unraid. I thought it was due to the drive but after swapping them it seems to be location based.

 

[audelair]

Yeah but I'm getting temperature above 40°c on one of the slot quite consistently. Which trigger the default warnings in unraid. I thought it was due to the drive but after swapping them it seems to be location based.

40°c seems awfully low for NVMe drives... maybe it's a setting more relevant to HDD's. Even with a heatsink, my NVMe goes from 38°c idle to 55°c under load.

If you're concerned though, they do have those heat sink blocks that are no wider than the M.2 drive, and are attached by rubber bands. So it barely adds to the width of the drive. I bet you can fit them.

 

[DaAliG]

My server resides in my living room, so I decided to go with the WD Red Plus drives. Compared to other (Pro) NAS drives (which are usually cheaper) these are fairly quiet. I also run the server with the HDD cover closed, I have not checked whether running it open has a huge impact on drive noise and temperature.

BTW, the DDR5-SO-DIMMs use on-die ECC, so possible RAM errors will be corrected. Unfortunately, information about the on-die ECC status is kept internally and cannot be accessed by the motherboard or the OS. But currently I think that running a good memory test once in a while is sufficient for me.

Running it with the cover off would indeed be unusual ha. I was trying to see if you had an idea of the sound dampening ability of the case. Open/closed = 50% softer or such. The only cheap drived in my region are enterprise and just concerned about noise. Paying more for lower capacity, "NAS" drives is always an option though....
Btw, for your cooler, how could you tell a particular cooler could be mounted? I'm also looking at getting the JONSBO radiator just so I have a guarenteed heatsink I could then add my own fan to. But your noctua isnt much more expensive in my area, so just asking how you were sure you could attach it right?

 

[chripo]

Running it with the cover off would indeed be unusual ha. I was trying to see if you had an idea of the sound dampening ability of the case. Open/closed = 50% softer or such. The only cheap drived in my region are enterprise and just concerned about noise. Paying more for lower capacity, "NAS" drives is always an option though....

I will probably check this later. I'm currently experimenting on ASPM/C-States/BIOS settings, so disks are temporarily disconnected.

Btw, for your cooler, how could you tell a particular cooler could be mounted? I'm also looking at getting the JONSBO radiator just so I have a guarenteed heatsink I could then add my own fan to. But your noctua isnt much more expensive in my area, so just asking how you were sure you could attach it right?

Any cooler compatible to Intel LGA115x/1200 (NH-L9i is) should basically fit on the board. FYI, the Noctua fans are quite powerful yet very silent (any HDD will produce more noise), but you could easily remove the fan and see whether you are satisfied with the temperatures you get when running just passive cooling.

 

[heromode]

All 8 cores are intel_pstate.


All 8 cores are powersave.

I don't know if I'm actually reaching C10 but I doubt it, based on calling cpupower frequency-info. Note the line indicating that the call to hardware failed. I'm guessing that's the problem but I haven't been able to figure out a fix.

EDIT - mix freq is 800, so no way I'm hitting any of the deeper C states. No idea why powertop and cpupower report differently.

analyzing CPU 0:
  driver: intel_pstate
  CPUs which run at the same hardware frequency: 0
  CPUs which need to have their frequency coordinated by software: 0
  maximum transition latency:  Cannot determine or is not supported.
  hardware limits: 800 MHz - 3.80 GHz
  available cpufreq governors: performance powersave
  current policy: frequency should be within 800 MHz and 800 MHz.
                  The governor "powersave" may decide which speed to use
                  within this range.
  current CPU frequency: Unable to call hardware
  current CPU frequency: 800 MHz (asserted by call to kernel)
  boost state support:
    Supported: yes
    Active: yes

hmm that seems wrong, 'frequency should be within 800 MHz and 800 MHz.' is wrong i think, should be 800 and 3.80.
Can you confirm your cores go to 3.8GHz using s-tui?

 

[chripo]

Today I experimented a bit with BIOS settings and ASPM (which by default is not available with this board).

tl;dr

Enabling and trying out ASPM and C-states on the i3-N305 board does not seem to help much in terms of power consumption. But since I am no BIOS/NUC/CWWK expert, YMMV. If you would like to reproduce my attempts or want to try out more yourself, here are the records of my journey.

Prerequisites

I was just running the board (i3-N305), the CPU cooler & fan (Noctua NH-L9i), the system fan (Noctua NF-A8 PWM) and one USB stick (for Unraid 6.12.6). The CPU fan is mostly off on idle loads (semi-passive) and the system fan constantly runs at a very low speed (speeding up just a little bit on heavy loads). I also set all CPU cores to use the "powersave" scaling governor using the following lines in my Unraid /boot/config/go file:

for i in {0..7}; do
echo powersave > /sys/devices/system/cpu/cpu${i}/cpufreq/scaling_governor
done

Step 1

I tried to enable all reasonable power saving options I could find in the BIOS (C-States -> enabled; System Agent PCIe root ports {1-3} -> ASPM L1 & Substates L1.1/L1.2; PCH-IO PCIe root ports {1-11} -> ASPM auto & Substates L1.1/1.2 & L1 Low enabled). But if I re-enter BIOS the PCH-IO settings are set back to "disabled" automatically (which I didn't notice at first). With the PCH-IO settings being disabled, I had no ASPM support (and consequently no low C states).

lspci -vv | awk '/ASPM/{print $0}' RS= | grep --color -P '(^[a-z0-9:.]+|ASPM )'
00:1c.0 PCI bridge: Intel Corporation Device 54ba (prog-if 00 [Normal decode])
                LnkCap: Port #3, Speed 8GT/s, Width x1, ASPM not supported
                LnkCtl: ASPM Disabled; RCB 64 bytes, Disabled- CommClk+
00:1c.3 PCI bridge: Intel Corporation Device 54bb (prog-if 00 [Normal decode])
                LnkCap: Port #4, Speed 8GT/s, Width x1, ASPM not supported
                LnkCtl: ASPM Disabled; RCB 64 bytes, Disabled- CommClk+
00:1c.6 PCI bridge: Intel Corporation Device 54be (prog-if 00 [Normal decode])
                LnkCap: Port #7, Speed 8GT/s, Width x1, ASPM not supported
                LnkCtl: ASPM Disabled; RCB 64 bytes, Disabled- CommClk+
00:1d.0 PCI bridge: Intel Corporation Device 54b0 (prog-if 00 [Normal decode])
                LnkCap: Port #9, Speed 8GT/s, Width x1, ASPM not supported
                LnkCtl: ASPM Disabled; RCB 64 bytes, Disabled- CommClk+
00:1d.2 PCI bridge: Intel Corporation Device 54b2 (prog-if 00 [Normal decode])
                LnkCap: Port #11, Speed 8GT/s, Width x1, ASPM not supported
                LnkCtl: ASPM Disabled; RCB 64 bytes, Disabled- CommClk-
01:00.0 Ethernet controller: Intel Corporation Ethernet Controller I226-V (rev 04)
                LnkCap: Port #0, Speed 5GT/s, Width x1, ASPM L1, Exit Latency L1 <4us
                LnkCtl: ASPM Disabled; RCB 64 bytes, Disabled- CommClk+
02:00.0 Ethernet controller: Intel Corporation Ethernet Controller I226-V (rev 04)
                LnkCap: Port #0, Speed 5GT/s, Width x1, ASPM L1, Exit Latency L1 <4us
                LnkCtl: ASPM Disabled; RCB 64 bytes, Disabled- CommClk+
03:00.0 Ethernet controller: Intel Corporation Ethernet Controller I226-V (rev 04)
                LnkCap: Port #0, Speed 5GT/s, Width x1, ASPM L1, Exit Latency L1 <4us
                LnkCtl: ASPM Disabled; RCB 64 bytes, Disabled- CommClk+
04:00.0 Ethernet controller: Intel Corporation Ethernet Controller I226-V (rev 04)
                LnkCap: Port #0, Speed 5GT/s, Width x1, ASPM L1, Exit Latency L1 <4us
                LnkCtl: ASPM Disabled; RCB 64 bytes, Disabled- CommClk+
05:00.0 SATA controller: JMicron Technology Corp. JMB58x AHCI SATA controller (prog-if 01 [AHCI 1.0])
                LnkCap: Port #0, Speed 8GT/s, Width x2, ASPM not supported
                LnkCtl: ASPM Disabled; RCB 64 bytes, Disabled- CommClk-

With these settings (which are more or less the CWWK defaults), my system draws about 18.7W from the wall on idle and about 39.2W on full load when running Prime95 v30.8 blend torture test.[1]

Step 2

I discovered that my test BIOS from CWWK shows several hidden information and settings by pressing CTRL+F1 inside the BIOS setup screen. I wouldn't be surprised if this is also possible with the original BIOS version. E.g. I could now set "Advanced -> RC ACPI Settings -> Native ASPM" to enabled, as well as some other options there. On idle, I measured about 17.7W with these settings. Full load stays about the same. So no big improvement (~1W) here, which very likely is within my measurement accuracy.

Step 3

I then manually set PCH-IO PCIe root ports {1-11} -> ASPM auto & Substates L1.1/1.2 and according to Unraid most devices now have ASPM enabled:

lspci -vv | awk '/ASPM/{print $0}' RS= | grep --color -P '(^[a-z0-9:.]+|ASPM )'
00:1c.0 PCI bridge: Intel Corporation Device 54ba (prog-if 00 [Normal decode])
                LnkCap: Port #3, Speed 8GT/s, Width x1, ASPM L0s L1, Exit Latency L0s <1us, L1 <16us
                LnkCtl: ASPM L1 Enabled; RCB 64 bytes, Disabled- CommClk+
00:1c.3 PCI bridge: Intel Corporation Device 54bb (prog-if 00 [Normal decode])
                LnkCap: Port #4, Speed 8GT/s, Width x1, ASPM L0s L1, Exit Latency L0s <1us, L1 <16us
                LnkCtl: ASPM L1 Enabled; RCB 64 bytes, Disabled- CommClk+
00:1c.6 PCI bridge: Intel Corporation Device 54be (prog-if 00 [Normal decode])
                LnkCap: Port #7, Speed 8GT/s, Width x1, ASPM L0s L1, Exit Latency L0s <1us, L1 <16us
                LnkCtl: ASPM L1 Enabled; RCB 64 bytes, Disabled- CommClk+
00:1d.0 PCI bridge: Intel Corporation Device 54b0 (prog-if 00 [Normal decode])
                LnkCap: Port #9, Speed 8GT/s, Width x1, ASPM L0s L1, Exit Latency L0s <1us, L1 <16us
                LnkCtl: ASPM L1 Enabled; RCB 64 bytes, Disabled- CommClk+
00:1d.2 PCI bridge: Intel Corporation Device 54b2 (prog-if 00 [Normal decode])
                LnkCap: Port #11, Speed 8GT/s, Width x1, ASPM L0s L1, Exit Latency L0s unlimited, L1 <16us
                LnkCtl: ASPM Disabled; RCB 64 bytes, Disabled- CommClk-
01:00.0 Ethernet controller: Intel Corporation Ethernet Controller I226-V (rev 04)
                LnkCap: Port #0, Speed 5GT/s, Width x1, ASPM L1, Exit Latency L1 <4us
                LnkCtl: ASPM L1 Enabled; RCB 64 bytes, Disabled- CommClk+
02:00.0 Ethernet controller: Intel Corporation Ethernet Controller I226-V (rev 04)
                LnkCap: Port #0, Speed 5GT/s, Width x1, ASPM L1, Exit Latency L1 <4us
                LnkCtl: ASPM L1 Enabled; RCB 64 bytes, Disabled- CommClk+
03:00.0 Ethernet controller: Intel Corporation Ethernet Controller I226-V (rev 04)
                LnkCap: Port #0, Speed 5GT/s, Width x1, ASPM L1, Exit Latency L1 <4us
                LnkCtl: ASPM L1 Enabled; RCB 64 bytes, Disabled- CommClk+
04:00.0 Ethernet controller: Intel Corporation Ethernet Controller I226-V (rev 04)
                LnkCap: Port #0, Speed 5GT/s, Width x1, ASPM L1, Exit Latency L1 <4us
                LnkCtl: ASPM L1 Enabled; RCB 64 bytes, Disabled- CommClk+
05:00.0 SATA controller: JMicron Technology Corp. JMB58x AHCI SATA controller (prog-if 01 [AHCI 1.0])
                LnkCap: Port #0, Speed 8GT/s, Width x2, ASPM not supported
                LnkCtl: ASPM Disabled; RCB 64 bytes, Disabled- CommClk-

With these settings, full load power consumption went down to about 35.9W (about ~3W less than stock) and idle to 17.0W. Again, not that big of an improvement and especially the little bit lower idle consumption is very likely within my measurement accuracy. But maybe the JMB585 SATA controller is preventing lower power consumption because of not supporting ASPM and therefore lower C states?

Step 4

I completely disabled PCH-IO PCIe root port 11 (which is the JMB SATA controller on my board) in the BIOS and made sure that all other PCIe ports had ASPM enabled. This setting will render 5 of the 6 SATA ports useless, so I tested it more out of curiosity whether struggling with ASPM and C states is really worth it.

lspci -vv | awk '/ASPM/{print $0}' RS= | grep --color -P '(^[a-z0-9:.]+|ASPM )'
00:1c.0 PCI bridge: Intel Corporation Device 54ba (prog-if 00 [Normal decode])
                LnkCap: Port #3, Speed 8GT/s, Width x1, ASPM L0s L1, Exit Latency L0s <1us, L1 <16us
                LnkCtl: ASPM L1 Enabled; RCB 64 bytes, Disabled- CommClk+
00:1c.3 PCI bridge: Intel Corporation Device 54bb (prog-if 00 [Normal decode])
                LnkCap: Port #4, Speed 8GT/s, Width x1, ASPM L0s L1, Exit Latency L0s <1us, L1 <16us
                LnkCtl: ASPM L1 Enabled; RCB 64 bytes, Disabled- CommClk+
00:1c.6 PCI bridge: Intel Corporation Device 54be (prog-if 00 [Normal decode])
                LnkCap: Port #7, Speed 8GT/s, Width x1, ASPM L0s L1, Exit Latency L0s <1us, L1 <16us
                LnkCtl: ASPM L1 Enabled; RCB 64 bytes, Disabled- CommClk+
00:1d.0 PCI bridge: Intel Corporation Device 54b0 (prog-if 00 [Normal decode])
                LnkCap: Port #9, Speed 8GT/s, Width x1, ASPM L0s L1, Exit Latency L0s <1us, L1 <16us
                LnkCtl: ASPM L1 Enabled; RCB 64 bytes, Disabled- CommClk+
01:00.0 Ethernet controller: Intel Corporation Ethernet Controller I226-V (rev 04)
                LnkCap: Port #0, Speed 5GT/s, Width x1, ASPM L1, Exit Latency L1 <4us
                LnkCtl: ASPM L1 Enabled; RCB 64 bytes, Disabled- CommClk+
02:00.0 Ethernet controller: Intel Corporation Ethernet Controller I226-V (rev 04)
                LnkCap: Port #0, Speed 5GT/s, Width x1, ASPM L1, Exit Latency L1 <4us
                LnkCtl: ASPM L1 Enabled; RCB 64 bytes, Disabled- CommClk+
03:00.0 Ethernet controller: Intel Corporation Ethernet Controller I226-V (rev 04)
                LnkCap: Port #0, Speed 5GT/s, Width x1, ASPM L1, Exit Latency L1 <4us
                LnkCtl: ASPM L1 Enabled; RCB 64 bytes, Disabled- CommClk+
04:00.0 Ethernet controller: Intel Corporation Ethernet Controller I226-V (rev 04)
                LnkCap: Port #0, Speed 5GT/s, Width x1, ASPM L1, Exit Latency L1 <4us
                LnkCtl: ASPM L1 Enabled; RCB 64 bytes, Disabled- CommClk+

I expected lower power consumptions, but measured (even a little bit higher) 37.8W on full load with these settings. On idle, the lowest power consumption I measured was about 14.6W, but remember that the JMB SATA controller was disabled. I would guess that the ~3W difference simply results from that.

Last but not least I ran powertop --auto-tune with these settings but didn't get any better idle measurements.

Conclusion

All of my attempts to improve the idle or full load power consumption of the CWWK i3-N305 NAS board did not show any signs of success. Disabling the JMB585 SATA controller resulted in a ~3W better idle consumption, but at the cost of 5 SATA ports.

It might be an option to add an NVME M.2 to 6xSATA 3.0 adapter card with an ASM1166 chip (which should be capable of ASPM and therefore possibly lower CPU C states) and leave the JMB controller disabled. But the adapter card will also need power and that will likely be in the ballpark of the 3W that can be saved by disabling the JMB controller.

I hope some people here will find this useful. If you have any suggestions on BIOS settings to further cut down power consumption, please let me know.

Edit: With more or less standard BIOS settings (see Step 1) and 4 HHDs plus 1 SSD attached to SATA, idle power consumption of my build is about 21.8W (disks not spinning).

[1] All P95 measurements were recorded after running ~5 minutes to avoid possible errors due to CPU boost.

 

[bugmenot]

I have buy couple day's ago n100 board from bkhd on aliexpress, and i hope ill be able to test this bios settings to see how low n100 can go with unraid and 1 ssd and 4 hdd's. Ill report my findings when i get my bord in 10-20 day's

 

[MWindu]

I hope some people here will find this useful. If you have any suggestions on BIOS settings to further cut down power consumption, please let me know.

Although the ethernet controllers are showing ASPM L1 enabled, I can’t help but wonder if disabling three of them could net some power savings.

 

[bogdyro]

Hi all.
Just got the n100 BKHD board from AliExpress. My goal was to build a low power homelab with pfsense maybe in the future.
This board has enough sata ports and 2 nvme + the 4 2.5gb intel ports.
This is the official site.

BKHD Mini-ITX Motherboard Inside N100 processor 6 SATA slots 2 M.2 High-Speed Storage 1264 NAS MB

BKHD 1264 NAS motherboard with built-in Intel N100 processor, quad-core and quad-thread, equipped with 4 Intel 2.5G Ethernet ports, 6 SATA ports and 2 M.2 ports, suitable for network storage, big data network transmission, remote access and mini server

www.bkipc.com

There is a bios available for download but the motherboard already ships with this version.
Idle power consumption with only 1 nvme and no sata, 16gb ddr5 4800, connected to 1 Gbit port is 14w in unraid.
Not great. Messing with the bios settings doesn't do much apparently. Applying powertop autotune results in unraid crashing.
The big no go for me though is the fact that doing wake on Lan after sleep the system freezes (no boot sound, no signal on monitor, it just hangs). Don't know if it could be drivers or bios settings.
If it continues like this have to send it back.
Also there is no possibility to control the CPU fan. Stays at max speed regardless of temps.
Nice hardware I guess but software or support will be 0

 

[MWindu]

Does the linux kernel in Unraid have full support for Alder Lake-N power management? The current Unraid release uses kernel version 6.1.64 yet I have found articles such as this that state kernel version 6.2 is needed for improved power management.

Could these tests be performed with Ubuntu or something with the latest linux kernel to see whether idle power consumption is decreased and wake on lan functions correctly?