CWWK/Topton/... Nxxx quad NIC router

[rufik]

Ask and you shall receive.......... the user manual for x86-P5 aka CW-X86-P5 aka CW-ADLNT-1C2L, useful especially for pin-outs and jumpers
CW-X86-P5-Chinese.pdf - original pdf file
CW-X86-P5-English.pdf - translated from Chinese to English with Google Translate

Could you re-upload manual please? This link is no longer valid...

 

[Drabon]

X86-P5 machine translated manual

 

[rufik]

I am back with more news and this time, good news.

At least with CW-AL v2 but I am sure that works with the others. We can compile a module to get info with pmw.

First, do a backup, things can always go wrong.

TROUBLESHOOTING AT THE END

First, leave your fan in your bios as automatic if you just one to see RPM or leave it as manual if you want to controle it in userland (AKA ourselves in proxmox)

# uname -r

Write down the version. In my case, 6.5.11-7-pve and then :

# apt-get install linux-headers-6.5.11-7-pve

Then clone this repo or copy its files with `scp` from your computer. Inside its folder do:

# make
# make install
# modprobe -a it87

Now we can use `sensors` to see if we get our new sensor. If not, first check that it is loaded:

# lsmod | grep it87

Example of my `sensors`



If everything works, add the module in your /etc/modules for autoload:

# echo "it87" >> /etc/modules

Now install `fancontrol` and execute:

# pwmconfig

If your fan is in manual, it will let you know which hwmon is it and it let you decide which piece of hardware you want as "driver" (in my case nvme) and then decide min temp, max temp, speed... Once done, it will be saved to /etc/fancontrol and that is it.

Works like a charm. Proper PWM, proper control and it is all you need. Happy computer.

I am happy I pushed this. I started with 54-58C NVME and now it is 45 and the FAN that I cannot hear at all. I can put it on the side to lose another 10 degrees if I need to.

After compiling it87 module from GitHub - frankcrawford/it87 it loads properly and fan control is being detected, thanks. But I cannot control 4pin PWM fan (neither from old PC chassis nor from dell laptop), pwmcontrol says:

Found the following devices:
   hwmon0 is acpitz
   hwmon1 is nvme
   hwmon2 is coretemp
   hwmon3 is it8613
Found the following PWM controls:
   hwmon3/pwm2           current value: 85
   hwmon3/pwm3           current value: 85

Giving the fans some time to reach full speed...
Found the following fan sensors:
   hwmon3/fan2_input     current speed: 0 ... skipping!
   hwmon3/fan3_input     current speed: 0 ... skipping!
There are no working fan sensors, all readings are 0.
Make sure you have a 3-wire fan connected.

I suppose CPU_FAN is `hwmon3/pwm2`, right?

Should the fan be standard 12V PWM 4-pin?

My board is CW-X86-P6-V1-N150 with 2 ETH ports. I think it's similar to CW-X86-P5...

 

[Drabon]

Are you trying to control the fans via software? My setup is controlled by the BIOS, but there is an option to set it to software in the fan settings.

And yes the Noctuas are 12V.

 

[splifingate]

Hah, indeed! But it seems that most likely the QEMU package update fixed the higher load issue.

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

I give you that (especially since I'm not really that Current on my own Sys

Are you still seeing improved Load?

 

[rufik]

Are you trying to control the fans via software? My setup is controlled by the BIOS, but there is an option to set it to software in the fan settings.
And yes the Noctuas are 12V.

Yes, I'm just trying to run fan anywise - through BIOS or software.
Is that pinout correct for CW-X86-P5?

edit: according to manual the fan connector has 1.25 pin spacing, so it's probably JST GH type.

 

[Drabon]

yes that looks correct. Though I would double check with a multimeter, since you have to solder the adapter anyway.

Edit: The Numbers "1" and "4" are even printed on the board (but hardly visible if you don't take off the cooling unit)

 

[xealous]

Just got a Topton n150 version. It's moderately warm to touch (installed OpenWRT on it). I can see some stuff about enabling PL1/PL2 (my bios has basically nothing to tune...). https://www.aliexpress.com/item/100....order_list.order_list_main.51.55d21802vO6neE

Is there an easy way to disable the speaker? It's a jumpscare at boot.

I regret getting topton after reading about it in some of the messages in the thread.

I'm thinking of 3d printing a bottom or front to enable airflow (wouldn't want the board to randomly die from no airflow + high heat). Are there any files? I assume the new bottom would be slightly taller (15-20mm) and fit a slim fan.

I found this one: Mini PC Slim Case Un-fanlessized (80x80x10mm fan mount) by r0bb10 | Download free STL model | Printables.com

 

[JTech]

I'm from the US and considering ordering from AliExpress before the tariffs take place. Does anyone know if tariffs are imposed based on purchase date or shipped date? In my experience it takes topton a few months to ship

 

[Drabon]

not from the us, not a lawyer, but speaking from pure logic i think only the import date is relevant (special cases like ivory should not apply here)

Edit: now that i remember an ... uhm (EU) friend of mine bought a device from CWWK directly and it seemed this friend didn't pay tariffs or taxes. So I guess CWWK has some ways...better go with Aliexpress. They seem to be more legal and probably have better warranty.

ymmw

 

[JurajSK]

Hello everyone, this thread is the spring of knowledge.
I have X86 P5 V3, Corsair 32GB DDR5 4800MHz CL40, SSD Vi3000 512GB, Noctua NF-B9 PWM, with zigbee dongle connected
running Proxmox, 1VM and 2 LXC for now, plan more...
I start with idle power 9,7W = SATA controller off, CPU governor max battery
today I add the ASPM settings according some previous posts user
name zeroflow

• Chipset > PCH-IO Config > PCI Express Configuration
  • DMI Link ASPM Control: Auto
  • Per Root Port: ASPM: Disabled -> Auto
• Chipset > SATA > Sata Controllers: Enabled -> Disabled
• Chipset > HD Audio > Enabled -> Disabled

• Chipset > System Agent > PCI Express
  • Per Root Port > ASPM: Disabled -> L0sL1

ended with 7+-watts

than I use this guide https://www.servethehome.com/how-to-pass-through-pcie-nics-with-proxmox-ve-on-intel-and-amd/
I think, it is because I disable iGPU while it is not connected to any VM,(this was done for preparing LibreELEC and HDMI output)
ended with 5,3W idle power

everything is working as you see energy spikes (compiling ESP yaml and som updates and reboots of VM)

I hope this will help reduce consumption... I am wondering if I will see BIOS tomorrow after disable iGPU

Edit: BIOS is visible

 

[jauling]

Just gave the 6.14 kernel a try in Proxmox, and immediately my dmesg got flooded by these errors:

# dmesg | grep igen6
[    3.775370] caller igen6_probe+0x1bc/0x8e0 [igen6_edac] mapping multiple BARs
[    3.775412] EDAC MC0: Giving out device to module igen6_edac controller Intel_client_SoC MC#0: DEV 0000:00:00.0 (POLLED)
[    3.775430] EDAC igen6 MC0: HANDLING IBECC MEMORY ERROR
[    3.775431] EDAC igen6 MC0: ADDR 0x7fffffffe0
[    3.775469] EDAC igen6: v2.5.1
[    4.824379] EDAC igen6 MC0: HANDLING IBECC MEMORY ERROR
[    4.824382] EDAC igen6 MC0: ADDR 0x7fffffffe0
[    5.852327] EDAC igen6 MC0: HANDLING IBECC MEMORY ERROR
[    5.852335] EDAC igen6 MC0: ADDR 0x7fffffffe0
[    6.872532] EDAC igen6 MC0: HANDLING IBECC MEMORY ERROR
[    6.872536] EDAC igen6 MC0: ADDR 0x7fffffffe0
[    7.896351] EDAC igen6 MC0: HANDLING IBECC MEMORY ERROR
[    7.896355] EDAC igen6 MC0: ADDR 0x7fffffffe0
[    8.920317] EDAC igen6 MC0: HANDLING IBECC MEMORY ERROR
[    8.920320] EDAC igen6 MC0: ADDR 0x7fffffffe0

I've got the long fin 4-port N100 CWWK with 16GB Corsair Vengeance CMSX16GX5M1A4800C40. When on 6.8.12-9-pve kernel, zero memory errors even though both kernels load the igen6_edac module. Another user states that he has no memory errors on this same 6.14 kernel, so I'm a bit worried. I had done (what I had thought was) extensive memtest86 testing with complete success.

Has anyone else tried this newish 6.14.0-1-pve kernel?

 

[crashtest]

Does anyone have an english manual for the original topic of this thread? I have one of the topton n100s(I think the model is CW-AD4L-N v1, might be v2) in the review that this thread links from but the only manual references I can find lead to dead links.

Long story short, I've had this thing for a while and am finally getting around to using it in the homelab. In particular, I'm looking for the pin outs on all the non-standard connectors on the mobo so I can, at the least, make my own adapters. It'd be nice to verify which connectors are what as well. I get the feeling the picture on aliexpress for these might not be exactly accurate.

 

[Drabon]

CWWK has some manuals in Chinese. Do you mean these?

CW-ADLN-4L用户手册2024.03.14更新 en

第12代AlderLake-N先锋四网-N100-N200-I3-N305-V2旗舰款规格书 en.pdf


CW-X86-P5_V2-V3用户手册2024.05.23更新 <-- this is just the updated x86-P5, all files translated with deepL

 

[pinako]

Well, this has been a roller coaster ride for me. I have an older N5100-based mini PC running OPNsense reliably for a couple of years now. I desired an upgrade, and TopTon was selling N150 boxes for the same price -- amazing! So I bought one... actually two. They came with H30W-N150-226 motherboard and AMI Aptio 2.22.1293 UEFI/BIOS (BK-1264NP-N150 Ver: 41.5 from 12/15/2024). They work, they boot into OPNsense, and they have not seen any driver support issues. But I could only get their idle power consumption down to 9-11 W.

Key lessons learned:

• The latest and greatest CPU might be marginally more efficient, but you might end up paying the difference (or more) for other components.
• Your router probably doesn't need PCIe Gen 4 x4 storage. Heck, this one wouldn't go faster than Gen 3 x1. Even SATA (or USB stick, microSD card, etc) would work just fine as a boot disk.
• Your router probably doesn't need DDR5. If you're asking if you should get the 5600 MHz RAM instead of the 4800 MHz RAM, be careful what you ask for. Alas, I can't just put some DDR4 into this one.
• If you're using this as a high-performance server, you're going to want a good cooling solution.
• If you just have a burning desire to upgrade your heating system, a little space heater or dutch oven would do the job for a fraction of the cost.
• Consult the STH forums before buying, not after. Yes, I did speed-read the 2854 posts in these 143 pages... thank you for all your input

The problem with these fanless N150 boxes from TopTon is they run hot. I thought the N5100 boxes were hot, but these new units broke the record for me. And yes, I get that the case is supposed to get hot, because that's its job, but I consider it a symptom of excessive power draw. Anyway, the BIOS is very limited. I was able to set `ASPM=auto`, but there were no power limit settings. In OPNsense, I set tunables `hw.acpi.cpu.cx_lowest=Cmax` for C-states and `dev.hwpstate_intel.[0-4].epp=100` for P-states as usual. Doing nothing, it burns 11-13 W and radiates 52-56 °C from the CPU. With 2 NICs active, it burns >13 W. In comparison, my older N5100 box pulls 9-10 W in production with CPU temperatures of 48-50 °C, on a workload that includes IDS monitoring of other network gear. And... the famous $88 Acer with the J4125 is still going strong on only 8-9 W.

Back in the very limited BIOS, I disabled SATA, serial ports, TPM, and 3 of 4 CPU cores... without making a dent in the power draw. The TopTon representative suggested putting a fan on it (LOL). They then shared that the CPU could be operated in 10W mode, 15W mode, or 25W mode with different performance characteristics, but did not offer any suggestions for actually making it happen. Finally, they admitted their BIOS is not actually able to set power limits. So I did the next best thing... which is to increase the "Tcc Activation Offset" to force thermal throttling. That actually locked all cores to 200 MHz and slowed the machine down to a crawl. CPU temperatures finally dipped below 50 °C, but there was only a ~0.1 W reduction of power draw. Not worth the hassle.

At this point, I realized that I'd done as much as I could to the CPU. Maybe the motherboard (or components) or PSU was the culprit. Well, I'd installed 16 GB of Mushkin Enhanced Redline DDR5-4800 40-40-40 and 256 GB of SK hynix BC901 PCIe Gen 4 x4 NVMe. Having run out of other options, I appropriated 128 GB of (cheap!) Patriot P320 PCIe Gen 3 x4 NVMe out of another server and put it into this thing. Just like that, the idle power draw decreased to 9-11 W and CPU temperatures decreased to 48-52 °C. A 2 W difference might not sound like much, but it's a significant improvement when the target is 9-10 W.

I have the same concerns as @phil-2024. There's something wrong about these purported low-power devices idling at >10 W, and it probably comes down to motherboard design and user choices such as RAM and SSD. My units also draw ~2 W when they're "off," and I chalked it up to the network ports and some USB ports being powered at all times. The cheap PSU by itself wastes 0.25 W plugged into nothing. As for longevity... my N5100 boxes have been simmering at close to 50 °C continuously for years now, so apparently it's fine. I might redo the thermal paste or add a shim, but the heat sink case does get rather toasty as is, so it's conducting reasonably well.

I see that @slybunda prefers Windows... and I did try Windows 10 & 11 on these things, but I haven't seen any better idle efficiency. Others have all sorts of elaborate setups involving Proxmox/virtualization to take advantage of Linux hardware support and PowerTOP magic, but I personally prefer to keep the configuration simple, so I have not yet worked up the courage to use a forbidden router.

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.

I mean... that's exactly why I have 2 of them, and they still cost less than Protectli. I also have the 2x N5100/i225 routers that are still functional, and the OPNsense config file appears to be compatible with all 4 devices without modification. They might not be suitable for mission-critical applications where downtime is unacceptable. But for home use, if you don't mind manually swapping in a spare, this might be fine

Wondering what people use all the ports for, One for WAN, one for LAN I guess, but if the LAN is going to a switch why the need of the other ports?

Not sure why you'd ever want to use more than 2 ports. If this is a router/firewall, you'd have 1 WAN and 1 LAN port. Leave switching to a real switch which has an ASIC designed for that.

For those wondering why we'd want more than 2 ports on a firewall/router, I have 2 words: management interface

Basically as per experts recommendations, you should run Suricata (intrusion protection) on WAN and Zenarmor on LAN (virus/malware protection), Zenarmor supposed to stop security risks spreading on Your LAN between devices...

i can see big businesses having a need for this, but the average home user.. im not convinced. not heard of anyone with stock isp router getting hacked. usually its something dodgy they downloaded that gets them got.

Okay, picture this: You're an average home user who just purchased a sketchy mini PC from AliExpress. You want to let it loose on your network (or even control your network), but you're not sure if you should trust it yet. You've installed a trusted OS, but there might be malware baked into the firmware or hardware. What do you do? Why, of course you'd plug it into your other pfSense/OPNsense box (which you also got from AliExpress), on a quarantined VLAN, with all the IDS bells and whistles, so you could keep an eye on it for a while!

 

[crashtest]

CWWK has some manuals in Chinese. Do you mean these?

CW-ADLN-4L用户手册2024.03.14更新 en

第12代AlderLake-N先锋四网-N100-N200-I3-N305-V2旗舰款规格书 en.pdf


CW-X86-P5_V2-V3用户手册2024.05.23更新 <-- this is just the updated x86-P5, all files translated with deepL

YES!!! THANK YOU!

 

[phil-2024]

Well, this has been a roller coaster ride for me. I have an older N5100-based mini PC running OPNsense reliably for a couple of years now. I desired an upgrade, and TopTon was selling N150 boxes for the same price -- amazing! So I bought one... actually two. They came with H30W-N150-226 motherboard and AMI Aptio 2.22.1293 UEFI/BIOS (BK-1264NP-N150 Ver: 41.5 from 12/15/2024). They work, they boot into OPNsense, and they have not seen any driver support issues. But I could only get their idle power consumption down to 9-11 W.

I've recently tried a Protectli VP4630, a step up in quality and substantial heatsink compared to the usual black finned devices from Aliexpress. I put on pfSense, and the idle power consumption was 12 watt like every other device in this class as no C-States enabled So I head into the AMI bios, no options for setting CPU C or P-States, that menu is missing! These devices allow swapping the BIOS to Coreboot, although it needs a USB stick with a live Ubuntu image to boot into, not very straight forward, why they can't use a very simple UEFI image like everyone else I don't know. So Coreboot has zero options to configure, but had been compiled to allow C-States, so power consumption dropped down to 6 watt without doing anything. Running Windows and Throttlestop to see what C-States it was dropping into, showed C-States were down to C8, however no P-States were enabled. I think if P-States were enabled we'd be looking at 4 watts maybe less, which equals my i7 7500 network appliance.

I've raised it with their tech support and they are speaking to their BIOS team about it (possibly that is Yanling in China), but it shows very little testing with these devices regarding power consumption takes place. If they tested properly and validated power consumption they'd notice a 50% increase in idle consumption with the AMI bios compared to Coreboot and would have looked into it. When running on a UPS, which they do test and show results for, you'd get double the run time with the Coreboot BIOS (assuming a lightly loaded box), which isn't insignificant, yet seemingly no one has spotted the differences and thought to see what was wrong with the AMI bios.

I used a UEFI Editor on the Protectli AMI bios and can see the CPU Power menu has defaults for C-States and P-States set to disabled. Intel don't recommend disabling C-States and don't make any guarantees to longevity if C-States are disabled.

The question is, why are the CPU Power options on these "out of China" type network appliances often hidden with C-States disabled that the user can't change? It's not because the menus are complicated, they leave far more complex and dangerous settings available. I'm getting suspicious that they are buying Intel SoCs that have failed testing when using lower power states, so no good for laptops or tablets, so get sold on for use in devices that are mains powered only. Even when we get access to the Power settings, as you have found, turning on C-States often doesn't make any difference anyway*, almost as though they are always hard coded as off

My 6 port N100 appliance is showing signs of being unstable now, it has corrupted data on the disk at least twice needing a full reinstall, and also when installing pfSense, it takes several attempts as CRC checks keep failing when it copies over the image, I've tried different USB sticks, and the corruption has happened going to a SATA drive and an NVMe drive.

They then shared that the CPU could be operated in 10W mode, 15W mode, or 25W mode with different performance characteristics, but did not offer any suggestions for actually making it happen.

So these modes are related to thermal designs and how much power it can use and for how long, these options don't do anything for idle power consumption as you have found.

As I have a 6 port i7 7500U appliance, this is a CPU that is 9 years old, with 3 active network connections currently idling at 4 watt, there is no way that a modern N100 or similar CPU in the same network type of appliance should be idling at 12 watts, something is very wrong somewhere. My Intel 10th generation laptop with the screen on a medium brightness only draws 10 watts, and with the screen off drops down to 3 watts.

* Worth noting that pfSense and OPNsense on FreeBSD, may show C-States are being entered, i.e. time spent at C1 or above, but this isn't what it seems. Those C-States reported are the OS suggestions for possible C-States due to current loads the OS is experiencing, it isn't saying the CPU has actually taken any notice and entered those states, and as power consumption doesn't change, then we can be 100% sure the CPU hasn't entered any C-States.

 

[Stovar]

https://www.notebookcheck.net/Low-power-Zen-6-CPU-cores-reportedly-beat-Intel-E-cores-in-power-consumption-while-bringing-75-IPC-of-Zen-5.994826.0.html

AMD Zen 5 LP cores power consumption
Per one of AMD’s internal slides shared by MLID, the Zen LP cores are “Client” products that target “Peak Core Efficiency”. The slide depicts the Zen 5 LP cores using 1 W or less per core, making the cores more efficient than Intel’s E-cores.


Hopefully AMD can bring a worthy competition to these N100/300 cpus, as long as they put more pci-e lanes and with good gpu acceleration, they can best them.

 

[douteiful]

I'm running a rackmount router with a Topton (BKHD 1264N4L) N100 board + OPNsense and it constantly draws 17W at idle, which I feel is too much. I tried changing the PL 1/2 to 6000/12000, disabling PECI, disabling Turbo, using C3 states, using powerd, and literally nothing seems to make a difference in the power draw, it stays at 17W. Am I doing something wrong?

A similar board on Linux draws 14W but also doesn't seem to change at all when I change the BIOS settings.

Note that I'm enabing a "Yellow screen fix" option in the BIOS to enable the rest of the power-related options. Maybe all of these "unlocked" options just do nothing?

 

[phil-2024]

I'm running a rackmount router with a Topton (BKHD 1264N4L) N100 board + OPNsense and it constantly draws 17W at idle, which I feel is too much. I tried changing the PL 1/2 to 6000/12000, disabling PECI, disabling Turbo, using C3 states, using powerd, and literally nothing seems to make a difference in the power draw, it stays at 17W. Am I doing something wrong?

A similar board on Linux draws 14W but also doesn't seem to change at all when I change the BIOS settings.

Note that I'm enabing a "Yellow screen fix" option in the BIOS to enable the rest of the power-related options. Maybe all of these "unlocked" options just do nothing?

Welcome to the club The PL settings only deal with the maximums and how much power can be drawn and for how long, they don't affect idle power consumption.

I have an older i7 7500U 6 LAN port which draws 4 watts at idle with 3 active network connections, so anything more modern built on more efficient lithography should not be drawing what we typically see when sat at idle.

I think you are right, unlocked settings do not change how these types of board work with power management and C-States, I've certainly seen no differences. So we can only speculate at what is going on. These devices are cheap, and cheap for a reason. Often they cost little more or even sometimes less than the cost of the Intel SoC they house, so we have to be suspicious.

So what could be going on? Well, it could be the AMI BIOS isn't coming with full support for these newer CPUs, and it wouldn't surprise me if the AMI BIOS that arrives on these cheap appliances is unlicensed and outdated. Would Chinese manufacturers ship with an unlicenced out of date BIOS, you bet they would if they can get away with it! For example, Topton advertise the ability as an OEM to supply their PCs with cracked Windows software!


Source Toptonpc

Building a motherboard with voltage regulation that can cope with sudden demands for power one millisecond, then dropping to next to nothing a few milliseconds later is expensive, it needs lots of testing and validation, extra components and good quality ones at that, have they just gone cheap on the voltage regulation then disabled power management options so the board remains stable?

The Intel SoCs? Why do we see the same SoCs flooding Aliexpress over the years? The N100 came from nowhere than is everywhere on Aliexpress, with no hardware seemingly able to use less than 12 watts when idle on a chip marketed by Intel for being ultra low power? Yet hardware from other vendors with an N100 can idle at a few watts, whether that is in branded laptops or single board computers from companies like Odroid.

For a long time it was flawed Celeron's in these types of devices being sold too cheap to be true, they never told customers it was because the CPUs were prone to failing over time due to a hardware flaw Intel update on Celeron J1800, J1900, N2807, N2930. Did these type of appliances ever get to ship with fixed steppings of Celeron's or just used to off load bad batches of CPUs? So are Intel selling SoC's that maybe failed validation for low power use (i.e. they aren't stable in low power states) so no good for their target audience of laptops and tablets, so they've been bought up cheap by Chinese factories that hide behind various Aliexpress brands and put to use in mains only devices like this and mini PCs?

A few years ago no one would be too worried about idle power consumption, although perhaps they were indirectly by trying to manage CPU temperatures, which is directly affected by C-States and Package C-States as if they aren't working or not working optimally, more power is used and more heat is generated. Now given the costs some of us pay for electricity, we are taking an interest in idle power consumption.

The other issue with running these chips with C-States disabled is Intel do not guarantee their CPUs for longevity, they are designed to have C-States being entered, so we are potentially running these chips 24/7 with no C-States being entered just to have them fail early, but then that's built in obsolescence and no one seems to expect these cheap devices to last more than a couple of years anyway.

If any representative from these Chinese factories are reading this, now is your chance to set the record straight and explain what is going on.