CWWK/Topton/... Nxxx quad NIC router

[spiffinite]

I am curious if my temps are in normal range. Cpu cores idle around 40C and nvme around 48C. With command "stress -v -c 4 -m 10" it causes all 4 cpu cores to peg 100% and temps go to 63C on the cpu. Nvme isnt being stressed with this.

 

[atomek]

Hello Everyone!

I plan to buy the N100 4 port platform, to run firewall (pfSense) and Gentoo in virtualisation. As it will run 24/7 it is very important for me to keep as low idle power consumption as possible. Could you give me some recommendations?

1) Which brand to chose? So far this one looks best for me https://aliexpress.com/item/1005005411300587.html (CWWK) - but i've heard some people got version with Y-shaped fins. How to make sure they will ship straight fins? Are any other brands better?

2) For virtualisation is it better to run Proxmox or ESXi (for low idle power and efficiency)

3) Does amount of RAM has impact on idle power usage? Are there specific BIOS settings to get the power usage down?

4) Could you recommend a wifi card for it and a good antena?

 

[wallysimmonds]

Started going down the rabbit hole on these after submitting the order for a gigabit service here in AU and looking for USG replacement options. Ubiquiti aren't lighting a fire under me with options so very interested in the roll your own option.

Looking at the CWWK options as Topton and Kingnovy don't seem to be rated as well here - I'm presuming the Pioneer edition is to be avoided due to the shorter fins?

 

[spiffinite]

Started going down the rabbit hole on these after submitting the order for a gigabit service here in AU and looking for USG replacement options. Ubiquiti aren't lighting a fire under me with options so very interested in the roll your own option.

Looking at the CWWK options as Topton and Kingnovy don't seem to be rated as well here - I'm presuming the Pioneer edition is to be avoided due to the shorter fins?

I can't fully comment yet as I'm testing my device, but the HUNSN RJ44 seems pretty good so far. The sell the CWWK/Topton devices on Amazon and you can return if things don't work out. Not that I'm hoping that will happen. I've got posts over the last few days on my testing so far. I haven't seen it reboot even after hours of stress testing after tweaking a few BIOS options.

They have pretty good support for a Chinese company and will provide spec sheets and BIOS updates if need be.

 

[alexw1982]

I just received the 4x NVME m.2 adapter board. Does anybody know what the switches JC3 and JC4 do?

 

[ToniCipriani]

I think the fan kit - or at least the cable should be included with all of these products, the SSD/Memory side has barely any airflow and just heats up.

edit: also I hear virtualised in proxmox runs cooler but I prefered to run bare metal

They do, actually, in China.

I've read they pretty much come standard on units sold on Taobao.

 

[alexw1982]

Installed the fan kit tonight with both NVMe M.2 slots utilized. The ssd that was in NVMe slot adapter for 2232 port was getting in contact with fan. I had to file down the screw mount on the adapter to 5mm in length to eliminate the contact … not pretty but it seems to work

during memtest the ram temperature max is now 56C (without fan in the low 80ies)

 

[ToniCipriani]

I am curious if my temps are in normal range. Cpu cores idle around 40C and nvme around 48C. With command "stress -v -c 4 -m 10" it causes all 4 cpu cores to peg 100% and temps go to 63C on the cpu. Nvme isnt being stressed with this.

N-series Tjmax is up to 105C, so I wouldn't worry too much.

My N305 has been running all day and it also hovers around that:

 

[terencelzw]

N-series Tjmax is up to 105C, so I wouldn't worry too much.

My N305 has been running all day and it also hovers around that:

View attachment 31675

Hey @ToniCipriani - are these temperatures at idle? I have a N305 6 port as well and have pretty similar temps as well (mid 50s to 60s) at idle. Unsure whether i'll need a fan to cool down the machine. It's definitely more than warm to the touch when i touch the case

 

[miken92]

Temperature wise this has been my observations over the weekend with a HUNSN RJ35 (big fins) N100

OPNsense bare metal - no fan
CPU fluctuating 50-57C mostly idle
SSD (Samsung 950 Pro) 59-62C
The old SSD runs quite hot and I suspect contributed to the overall temperature

OPNsense bare metal - fan
PWM min 50%, can't remember the exact settings but it did turn off so minimal fan speed
CPU 41C
SSD 52C

Proxmox + OPNsense (2 cores / 4GB / 2 bridged ports) - fan
Easily manages 1Gbps/1Gbps
Replaced SSD with Crucial P3
Tweaked fan options to be always on
CPU 35C (low to mid 40s at 1Gbps)
SSD 41C

coretemp-isa-0000​

Adapter: ISA adapter​

Package id 0: +35.0°C (high = +105.0°C, crit = +105.0°C)​

Core 0: +35.0°C (high = +105.0°C, crit = +105.0°C)​

Core 1: +35.0°C (high = +105.0°C, crit = +105.0°C)​

Core 2: +35.0°C (high = +105.0°C, crit = +105.0°C)​

Core 3: +35.0°C (high = +105.0°C, crit = +105.0°C)​

​

acpitz-acpi-0​

Adapter: ACPI interface​

temp1: +27.8°C (crit = +110.0°C)​

​

nvme-pci-0100​

Adapter: PCI adapter​

Composite: +40.9°C (low = -0.1°C, high = +84.8°C)​

(crit = +94.8°C)​

Sensor 1: +40.9°C (low = -273.1°C, high = +65261.8°C)​

Sensor 2: +49.9°C (low = -273.1°C, high = +65261.8°C)​

Sensor 8: +40.9°C (low = -273.1°C, high = +65261.8°C)​

The supplied fan does make a pretty annoying buzzing, but it is in a pretty noisy environment so not an issue for me and you'd have to quite close to hear it.

I'm much happier with the temperatures and the case is barely warm now.

 

[alexw1982]

has anyone been able to read the fan speeds within linux / proxmox. I failed to get it working using lm-sensors ...

 

[ToniCipriani]

Hey @ToniCipriani - are these temperatures at idle? I have a N305 6 port as well and have pretty similar temps as well (mid 50s to 60s) at idle. Unsure whether i'll need a fan to cool down the machine. It's definitely more than warm to the touch when i touch the case

Was configuring and testing a few things, the box has my OPNsense and Docker server on it running some containers. Maybe around 2-30%?

But "warm to touch" just means the heatsink is working IMO, I wouldn't be too worried about it. Point is the heat is not just on the chip cooking it.

It's when the heatsink feels cool but the chip is cooking, that's when you should worry (i.e. the thermal material has failed).

 

[alstair]

Variation A
N200
ADLN 0.01 x64, 04/04/2023

>> but I can't get the second SSD running, it can be seen in the BIOS, but Proxmox doesn't see it.

I've only very briefly used Proxmox, but it looks like there's some manual steps required to make additional storage available in Proxmox - could this be your issue?

Proxmox add disk storage space - NVMe drive

Proxmox add disk storage space NVMe drive. A look at the process to add an additional NVMe drive to a Proxmox host for VMs and containers

www.virtualizationhowto.com

Otherwise I suggest contacting your seller, as apparently these boards caused some issues during development:

Did you ever find a solution?
I'm having the same issue - two NVMEs, both recognized in BIOS, but proxmox installer only sees one.

 

[AnthonyUK]

Your opnsense is running bsd. I don't think cake qos is available for bsd kernels. So we are not running the same workload. My box is rougly loaded at 40-50% when i max bandwidth in one direction. Opnsense != openwrt. Also other features can drastically can increase cpu load as well.

I'm sure you have your reasons for running OpenWRT but it appears that OPNSense is much more efficient here so as I said, you don't HAVE to. There are other options.

 

[alexw1982]

Did you ever find a solution?
I'm having the same issue - two NVMEs, both recognized in BIOS, but proxmox installer only sees one.

Try to disable ASPM Auto setting in the BIOS settings for the relevant PCIe Ports. That's what fixed it for me. In general, setting ASPM to auto for SSD PCIe ports produced errors in proxmox for me...

in general that 2230 wifi adapter board is a janky solution. I would wish for a 2x PCIe3.0 x2 adapter board for running Raid1 setups ...

 

[alstair]

Try to disable ASPM Auto setting in the BIOS settings for the relevant PCIe Ports. That's what fixed it for me. In general, setting ASPM to auto for SSD PCIe ports produced errors in proxmox for me...

in general that 2230 wifi adapter board is a janky solution. I would wish for a 2x PCIe3.0 x2 adapter board for running Raid1 setups ...

Thanks, I've checked and ASPM is disabled by default on all listed PCI ports in the bios.

I'm thinking it might be an issue with the installer itself now. I gave up and restarted the device to see what would happen and it booted from nvme0n1 and nvme1n1 is showing up as an unformatted disk in the proxmox UI.

 

[alexw1982]

Thanks, I've checked and ASPM is disabled by default on all listed PCI ports in the bios.

I'm thinking it might be an issue with the installer itself now. I gave up and restarted the device to see what would happen and it booted from nvme0n1 and nvme1n1 is showing up as an unformatted disk in the proxmox UI.

Strange - for me it definitely fixed it ...

maybe somebody can help me out with regards to PL1/PL2 limits. I understand the basic concept of what the result is of setting PL1 and PL2. My question is whether it actually makes sense to set a low PL1 (if cooling is sufficient). Shouldn't the total power usage (in terms of kWh) be same but the task is simply executed more quickly if I set a high PL1 limit (e.g. 15000 for the N100 - looking at the power usage anything beyond that does not seem to increase power consumption / performance). With the internal fan I am able to run PL1=15000 with the CPU staying around 55°C ...

 

[devast]

I'm sure you have your reasons for running OpenWRT but it appears that OPNSense is much more efficient here so as I said, you don't HAVE to. There are other options.

It's not more "efficient". Not being able to do cake sqm does not make it any better. I have my reasons why i must run linux kernel ( pppoe, cake sqm, wireguard, etc) not bsd. I don't have to run anything cpu intensive, sure, but then why buy strong cpu for a lot of $$$? My $50 er-x could easily do 1gbit nat.

 

[pigr8]

It's not more "efficient". Not being able to do cake sqm does not make it any better. I have my reasons why i must run linux kernel ( pppoe, cake sqm, wireguard, etc) not bsd. I don't have to run anything cpu intensive, sure, but then why buy strong cpu for a lot of $$$? My $50 er-x could easily do 1gbit nat.

i also have a pppoe connection (1000/300) and a n5105 handles it like a champ, openvpn and wireguard clients no problem and bufflerbloat can also be fixed





i would like to know what openwrt could do better to test it

i assume the new n100 (x86-p5) that is coming replacing this bad boi will be better, i will compare the results and post them here.

 

[fta]

i would like to know what openwrt could do better to test it

I don't use openwrt, but I do use linux because, like @devast, I want CAKE SQM. To get somewhat equivalent functionality to CAKE in opnsense, you'd have to do htb+fq_codel since FreeBSD does not have a CAKE implementation. I don't think you can configure that from the GUI. The GUI config for shaping in opnsense is, IMO, pretty limited. CAKE works great for fairness and shaping and also works with hardware offload. The disadvantage is it uses quite a bit more CPU than fq_codel (but less than htb+fq_codel).

Cake - Bufferbloat.net

www.bufferbloat.net