CWWK/Topton/... Nxxx quad NIC router

Notice: Page may contain affiliate links for which we may earn a small commission through services like Amazon Affiliates or Skimlinks.

phil-2024

Member
Sep 7, 2024
48
36
18
You're seeing C states on the cores, but none on the entire package. I think this is a common issue on the N series though we can't always prove if it's the cpu, board or bios.

When devices have had bioses with working ASPM and related settings, it has been possible to enable things like PCIe & sata power saving which has allowed package C states and reduced the idle wattage. That can be a magical black art though since any peripheral or circuit (or bios/driver element) that does not correctly handle the power down and restart will lead to crashes. We're luckiest when the element says "NO", since that won't introduce crashes. Some of the cwwk bioses have had the ASPM controls, but others have it blocked off because - we assume - too many people were setting it too aggressively for the hardware and creating failures and support calls.

Eg. NAS motherboards with the JMB sata chips struggle a lot as they don't appear to support ASPM well, or possibly at all. The purple/K board with ASM1166 instead was invented to avoid that issue but first (& some? later?) bioses still hid the settings.

It's also a clear thing that most N series devices fail to power down some sections of the motherboard leading to high-ish idle watts. That could just be package c states but feels like it's weak design in the bios/uefi drivers or actual hardware which keeps some elements at full power (eg. Network, SATA, usb).

I do wonder if the same might even be true of some branded NUCs & motherboards, but as they usually have 1x1gb network, 1-2 sata etc it may not show. It is clear that the earlier generations were better overall at this - which is sad.
I've managed to unlock ASPM settings in the BIOS, I did get all peripherals showing as ASPM enabled, this saved about 2 watts on idle, however the CPU was still not entering any sort of C state, it was saying almost 100% in C3 in pfSense but this resulted in no power savings. I could switch C-states on and off and see absolutely no change in power consumption. Like I've said I have an older i7 box with 6 Ethernet ports which with no difficulty at all is idling at ~4 watts, practically cold to the touch and just runs without issue, never ever crashed, so these appliances are easily able to run at very low power. It is very annoying that a newer more efficient processor in the same form factor is drawing at idle almost 3 times as much power, and even uses more power when it needs to ramp up. Its funny how all these N100 type appliances from China are all idling around the 11-12 watt.

I'm thinking of getting a Protectli VP6630 to get something branded with hopefully some support, would I see the same issues? In my case I will just put in a SATA drive which shouldn't really upset things.

There is also this interesting article here which suggests its a problem with all modern Intel processors: The Curse of Intel 12th Gen C-States
 
  • Like
Reactions: KevinR and Stovar

pigr8

Active Member
Jul 13, 2017
108
114
43
lol the VP6630 costs 5 times as much as the n100 box.
exactly, many do calculation about power consumption and extreme optimizing to squeeze out those ~3-4w but it makes little to no sense to spend 5x times the amount of those boxes to get a lower electric bill.. how many months/years have to pass to get even and start to save up something?

i get the point that they could consume less power and be more optimize but tbh if i have a box running at 11w instead of 7w for a months would save how much? here @0,25€ kWh a month would be 1.98€ for 11w and 1.26€ for 7w, this is a difference of 72 cents a month running 24/7.. so to me it would be smarter to get the whole picture.

and this assumes that branded boxes (with no real guarantee that would perform different or get better support) are actually better.

ok in a heavy packed homelab with 15 of those boxes running maybe it could make sense so it would not trip the breaker for exceeding power but even there the upfront cost would be huge for little to no save in constant use.

m2c
 
  • Like
Reactions: Stovar

phil-2024

Member
Sep 7, 2024
48
36
18
lol the VP6630 costs 5 times as much as the n100 box.
They are more expensive indeed than a N100 box from China, but the VP6630 is a more powerful processor and it has 2 SFP slots. My reasoning is, if I'm paying a premium for something 'branded' that will have a usable warranty and ongoing support with BIOS updates, I might as well invest in something that will last a length of time and is upgradeable. A similar spec'd equivalent to the VP6630 direct from China would cost me ~£425 but that is barebones so would need memory and storage, so lets call it £500, the VP6630 bare bones is €599 which is around £500, so lets call it £575 with memory and storage, and also comes with a real warranty, technical support and firmware updates. Someone would be silly to buy a similar spec'd box to the VP6630 from a no name Chinese seller to save £75.00!

exactly, many do calculation about power consumption and extreme optimizing to squeeze out those ~3-4w but it makes little to no sense to spend 5x times the amount of those boxes to get a lower electric bill.. how many months/years have to pass to get even and start to save up something?
It's not all about the cost of power, there are other considerations, such as cooling and longevity of the components when they bake 24/7, and run time on a UPS. Although in terms of power costs, my current box I've had for 7 years running 24/7 and at todays rates that would have cost at 12 watt over 7 years around £180, whereas it runs at 4 watt idle and so has cost me ~£60.00, the savings are actually much greater than that but hard to quantify exactly, but the current box is using much less power when under load compared to the N100. If it wasn't for the fact of needing 2.5G Ethernet ports to take advantage of faster Internet service, I'd still be using the same box and could see it going on for another 7 years.

I also wonder what is wrong with these boxes to need to draw so much power, especially given the N100 is suppose to be a low power chip, so how long will it last? I'm also frustrated with the lack of support from these Chinese companies, I emailed CWWK asking about BIOS updates and power consumption months ago and heard nothing. The BIOS will never get updated no matter what vulnerability is discovered. You get what you pay for.

Many of us except these types of kit from China are cheap for a reason and if they break after a year, we just buy another one, they are almost disposable, and we take a gamble when we buy them. That gamble for me hasn't paid off as this N100 appliance I don't feel confident in using, so I have to write off that expense and look for something else, so it doesn't always work out cheaper chancing what we get direct from China.
 
Last edited:
  • Like
Reactions: pinako

blunden

Well-Known Member
Nov 29, 2019
882
293
63
I'm also frustrated with the lack of support from these Chinese companies, I emailed CWWK asking about BIOS updates and power consumption months ago and heard nothing. The BIOS will never get updated no matter what vulnerability is discovered. You get what you pay for.
The lack of support is absolutely a valid concern. Protectli will provide that. Their hardware tend to be from one of these OEM manufacturers though, nothing special in that regard. That was always the case in the past at least. It's the support and firmware development that you pay extra for. :)

Regarding BIOS updates, the overwhelming majority of the security updates we see from BIOS updates come from microcode updates. Those can be applied at boot by most OS:es so you don't necessarily need BIOS updates to take advantage of most of those. Also, some of these OEM manufacturers do publish BIOS updates sometimes.
 

phil-2024

Member
Sep 7, 2024
48
36
18
The lack of support is absolutely a valid concern. Protectli will provide that. Their hardware tend to be from one of these OEM manufacturers though, nothing special in that regard. That was always the case in the past at least. It's the support and firmware development that you pay extra for. :)
Absolutely yes, the boards will be made by the same companies in China, and just assembled elsewhere to get a Made in the US label. At least with the VP66xx series they are not just rebadged Yanlings, the cases/cooling are unique to Protectli, and nowhere on Aliexpress or Alibaba are they any network appliances with the same arrangement of ports (showing that they likely contain the same board just in a different case), or the motherboard available on its own. So they seem a unique design for Protectli and I would hope a bit more time, care and testing went into them.

I really think in a lot of cases this kit we buy direct from China is just the Chinese manufacturers getting rid of bad batches of boards or using up components not fit for use in branded goods, its why the vast majority don't have any brand name on them, and we end up having to make modifications to get the CPU to touch the heatsink or make our own BIOS modifications and debug things ourselves :cool:
 
Last edited:

Stovar

Active Member
Dec 27, 2022
210
101
43
I did try and contact cwwk support a few times, when I had my P5 Nas board, but he appears to have stopped responding to emails or aliexpress cwwk chats and this was few months ago. I was just asking for an bios update also, so not exactly major support.

Hopefully cwwk support starts again but can't bank on it.
 

jauling

New Member
Jan 7, 2025
17
4
3
Amsterdam
I've got the long fin 4port CWWK case N100. Not sure why, but I powered off my server for a week while on holiday, and now I'm back. I upgraded Proxmox VE to 8.3.5, rebooted, and now for some reason it no longer idles at 700MHz like it did before. It's sitting at 1200MHz, and sometimes randomly one core registers as 700MHz. Very strange. Running kernel 6.8.12-8-pve

dmidecode says:
Code:
BIOS Information
        Vendor: American Megatrends International, LLC.
        Version: ALN4L102
        Release Date: 11/08/2023
        BIOS Revision: 5.26
And I've been doing these CPU scaling and performance settings on boot for a while now:
Code:
# crontab -l
@reboot (sleep 60 && echo powersave | tee /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor)
@reboot (sleep 65 && echo balance_power | tee /sys/devices/system/cpu/cpu*/cpufreq/energy_performance_preference)
 

blunden

Well-Known Member
Nov 29, 2019
882
293
63
Absolutely yes, the boards will be made by the same companies in China, and just assembled elsewhere to get a Made in the US label. At least with the VP66xx series they are not just rebadged Yanlings, the cases/cooling are unique to Protectli, and nowhere on Aliexpress or Alibaba are they any network appliances with the same arrangement of ports (showing that they likely contain the same board just in a different case), or the motherboard available on its own. So they seem a unique design for Protectli and I would hope a bit more time, care and testing went into them.

I really think in a lot of cases this kit we buy direct from China is just the Chinese manufacturers getting rid of bad batches of boards or using up components not fit for use in branded goods, its why the vast majority don't have any brand name on them, and we end up having to make modifications to get the CPU to touch the heatsink or make our own BIOS modifications and debug things ourselves :cool:
They might simply be made by an OEM you haven't looked at, or customized by one of the well known ones. A teardown would be interesting though. :)

That's just pure speculation on your part. :D It's more likely that the OEMs sell direct to consumers in addition to doing brand deals, especially if the brand isn't asking for hardware exclusivity.

It's possible that they don't care as much about QC on the ones sold to consumers though, who knows. Your argument about BIOS modifications being an indication of bad batches make no sense though as they'll run the same BIOS as the branded units, but with minor tweaks to update vendor strings at most.

Neither of my two Qotom boxes have the build quality issues some people mention with their CWWK boxes btw. so perhaps it's just a matter of OEM quality. This is an admittedly very small sample size though, so who knows. :)
 

Mike9474593

Active Member
Aug 5, 2022
135
95
28
Neither of my two Qotom boxes have the build quality issues some people mention with their CWWK boxes btw. so perhaps it's just a matter of OEM quality. This is an admittedly very small sample size though, so who knows. :)
My Kaby Lake Celeron from Qotom was fine, while the Celeron J4125 from Qotom had a gap between CPU and cooler, had to add a copper shim, the 5th network port did not work reliable. There are also some issues around with the Qotom Denverton System. My CWWK unit (N6005, first board revision) was not so pleasant (hot as hell, no low C States even in linux, BIOS buggy), the CWWK N5105 V4 is fine and stable, a CWWK X86-P2 (J4125) is also fine. I think its a hit or miss with these units. Probably the quality standards are too lax, some units are good, some bad. R&D is probably also cut to a minimum.
 
Last edited:
  • Like
Reactions: phil-2024

blunden

Well-Known Member
Nov 29, 2019
882
293
63
My Kaby Lake Celeron from Qotom was fine, while the Celeron J4125 from Qotom had a gap between CPU and cooler, had to add a copper shim, the 5th network port did not work reliable. There are also some issues around with the Qotom Deverton System. My CWWK unit (N6005, first board revision) was not so pleasant (hot as hell, no low C States even in linux, BIOS buggy), the CWWK N5105 V4 is fine and stable, a CWWK X86-P2 (J4125) is also fine. I think its a hit or miss with these units. Probably the quality standards are too lax, some units are good, some bad. R&D is probably also cut to a minimum.
Perhaps I was just lucky twice then. That's certainly possible. :)

They probably don't do enough QC. For instance, I don't necessarily think they do any burn in testing. If you buy them with RAM and storage, they'll install an OS and presumably check that the hardware work. I don't know if they do anything else.

I don't think they intentionally offload "bad batches" on AliExpress though. It's more likely to simply be the luck of the draw. :)
 
  • Like
Reactions: Mike9474593

Mike9474593

Active Member
Aug 5, 2022
135
95
28
Perhaps I was just lucky twice then. That's certainly possible. :)
Or the quality has decreased in the last years significantly. My first Qotom was from somwhere in 2019, the other was from 2022 or 2023. I had the impression the older case was a better (copper block between CPU and case instead of aluminium, parts fitted better).

They probably don't do enough QC. For instance, I don't necessarily think they do any burn in testing. If you buy them with RAM and storage, they'll install an OS and presumably check that the hardware work. I don't know if they do anything else.
QC cost money, and these units are optimized for low prices. Make your guess... :)

I don't think they intentionally offload "bad batches" on AliExpress though. It's more likely to simply be the luck of the draw. :)
Narrow tolerances are also a price question, so it's a lot cheaper if you can just produce the mechanical parts with wide tolerances, no hand selected chips and just a a default bios from AMI with almost no customization.
I don't think they do separate good from bad units intentionally, the production is just cost-optimized, everything that costs money is minimized or left out. This means from good to bad everything is possible... Sometime you win with a good unit, sometime you loose with a bad one.
 
  • Like
Reactions: blunden

blunden

Well-Known Member
Nov 29, 2019
882
293
63
Or the quality has decreased in the last years significantly. My first Qotom was from somwhere in 2019, the other was from 2022 or 2023. I had the impression the older case was a better (copper block between CPU and case instead of aluminium, parts fitted better).
My second unit was from earlier this month and it's perfectly fine. :)

I agree with the rest of your post. :)
 
  • Like
Reactions: Mike9474593

Mike9474593

Active Member
Aug 5, 2022
135
95
28
My second unit was from earlier this month and it's perfectly fine. :)

I agree with the rest of your post. :)
Which one did you get? I'm a bit under impression Qotom is falling a bit behind technology wise, mSATA instead of M.2, still using i225 instead of i226, no recent CPU. But I could be wrong. Unhappily they went up with shipping rates to my place, so their units are a bit too expensive compared to the CWWK ones.
 

blunden

Well-Known Member
Nov 29, 2019
882
293
63
Which one did you get? I'm a bit under impression Qotom is falling a bit behind technology wise, mSATA instead of M.2, still using i225 instead of i226, no recent CPU. But I could be wrong. Unhappily they went up with shipping rates to my place, so their units are a bit too expensive compared to the CWWK ones.
I bought the Qotom Q10922H6 N100 model and use it as a mini server. I also have the Qotom C3758 Denverton model as a router. Neither of those use mSATA and the newer one uses the i226V (but I use one of the 10 Gbps ports).

Yeah, they increased the shipping rates, but I thankfully was able to use a $40 coupon from an AliExpress sale at the time to lower the price a bit. The pricing difference wasn't that big compared to the CWWK boxes, and those models were also worse for my use case in some ways.
 
  • Like
Reactions: Mike9474593

Mike9474593

Active Member
Aug 5, 2022
135
95
28
I bought the Qotom Q10922H6 N100 model and use it as a mini server. I also have the Qotom C3758 Denverton model as a router. Neither of those use mSATA and the newer one uses the i226V (but I use one of the 10 Gbps ports).

Yeah, they increased the shipping rates, but I thankfully was able to use a $40 coupon from an AliExpress sale at the time to lower the price a bit. The pricing difference wasn't that big compared to the CWWK boxes, and those models were also worse for my use case in some ways.
Glad they caught up, different manufacturers are good for us customers.
The Q10900H6 S13 has great specs, I've already got my eye on it. It's ticking almost every checkbox on my list (the only thing I personally miss is a SFP+ port, I need one 10G copper and one SFP+ port). Unhappily there are no drivers for OPNsense for the AQC113, so I need to wait.
Glad the unit is fine!
 

blunden

Well-Known Member
Nov 29, 2019
882
293
63
Glad they caught up, different manufacturers are good for us customers.
The Q10900H6 S13 has great specs, I've already got my eye on it. It's ticking almost every checkbox on my list (the only thing I personally miss is a SFP+ port, I need one 10G copper and one SFP+ port). Unhappily there are no drivers for OPNsense for the AQC113, so I need to wait.
Glad the unit is fine!
Yeah, an SFP+ NIC would've been optimal for me as well, but I'm happy that it doesn't use the ancient Intel X520 that's about to go EoL. I run Linux on mine so drivers are not a problem.

As a router, pfSense and OPNsense seem to perform worse anyway if you try to use it as a router at 10 Gbps or greater.
 

jauling

New Member
Jan 7, 2025
17
4
3
Amsterdam
Welp, not sure which of these packages fixed it, but these just came out today, and after reboot I'm back to 700MHz.

Code:
novnc-pve/stable 1.6.0-2 all [upgradable from: 1.5.0-1]
pve-edk2-firmware-legacy/stable 4.2025.02-2 all [upgradable from: 4.2023.08-4]
pve-edk2-firmware-ovmf/stable 4.2025.02-2 all [upgradable from: 4.2023.08-4]
pve-edk2-firmware/stable 4.2025.02-2 all [upgradable from: 4.2023.08-4]
pve-firmware/stable 3.15-2 all [upgradable from: 3.14-3]
pve-qemu-kvm/stable 9.2.0-3 amd64 [upgradable from: 9.2.0-2]
I've got the long fin 4port CWWK case N100. Not sure why, but I powered off my server for a week while on holiday, and now I'm back. I upgraded Proxmox VE to 8.3.5, rebooted, and now for some reason it no longer idles at 700MHz like it did before. It's sitting at 1200MHz, and sometimes randomly one core registers as 700MHz. Very strange. Running kernel 6.8.12-8-pve

dmidecode says:
Code:
BIOS Information
        Vendor: American Megatrends International, LLC.
        Version: ALN4L102
        Release Date: 11/08/2023
        BIOS Revision: 5.26
And I've been doing these CPU scaling and performance settings on boot for a while now:
Code:
# crontab -l
@reboot (sleep 60 && echo powersave | tee /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor)
@reboot (sleep 65 && echo balance_power | tee /sys/devices/system/cpu/cpu*/cpufreq/energy_performance_preference)
 
  • Like
Reactions: splifingate

splifingate

Member
Oct 7, 2023
63
54
18
Welp, not sure which of these packages fixed it, but these just came out today, and after reboot I'm back to 700MHz.

Code:
novnc-pve/stable 1.6.0-2 all [upgradable from: 1.5.0-1]
pve-edk2-firmware-legacy/stable 4.2025.02-2 all [upgradable from: 4.2023.08-4]
pve-edk2-firmware-ovmf/stable 4.2025.02-2 all [upgradable from: 4.2023.08-4]
pve-edk2-firmware/stable 4.2025.02-2 all [upgradable from: 4.2023.08-4]
pve-firmware/stable 3.15-2 all [upgradable from: 3.14-3]
pve-qemu-kvm/stable 9.2.0-3 amd64 [upgradable from: 9.2.0-2]
A *lot* of stuff happened between 2023 and 2025! :)
 

jauling

New Member
Jan 7, 2025
17
4
3
Amsterdam
Hah, indeed! But it seems that most likely the QEMU package update fixed the higher load issue.

Code:
pve-qemu-kvm (9.2.0-3) bookworm; urgency=medium

  * revert changes to the High Precision Event Timer (HPET) to fix performance
    regression

 -- Proxmox Support Team <support@proxmox.com>  Wed, 26 Mar 2025 09:56:01 +0100
A *lot* of stuff happened between 2023 and 2025! :)
 
  • Like
Reactions: splifingate