CWWK/Topton/... Nxxx quad NIC router

[AnthonyUK]

Only issue with this is that it will shift the alignments of your ports with the endplates.

I very much doubt the tolerances are so close that removing a fraction of a mm will impact this.

 

[AnthonyUK]

acpitz-acpi-0
Adapter: ACPI interface
temp1: +27.8°C (crit = +110.0°C)

Does anyone know what this is as it is the same on every similar box it appears?

Mine as another example
acpitz-acpi-0
Adapter: ACPI interface
temp1: +27.8°C (crit = +110.0°C)

 

[shanestorey]

Has somebody tried to put a thermal pad between the heatsink and CPU to close the gap? I reapplied the thermal paste (but not between the copper block and heatsink). Temps are about 54ºC in idle.

I added a fan and it dropped only couple of degrees, still 50ºC. I tried to change the fan speed in BIOS, but it does not do anything, so the fan is spinning pretty low RPM. (I installed the fan in to the fan header with the adapter)

Edit. And for reference CWWK 6-port model

I've got a Topton N100 4 port. When it arrived, I cleaned off the thermal paste and could see quite a gap between the CPU and the copper block.

I ended up shimming the gap using aluminum foil between the chassis and copper block. I reasoned that foil would be thermally good enough and easy to find compared to copper.

Took a few tries to get the gap closed up, from memory it was six layers of foil with a dab of thermal paste between each layer.

It is installed in a cupboard that's at a temperature of 22c, running OPNSense native, so not a great load... in fact it pretty much idles all the time.
Temperature reported fanless was 39c and at the moment with a small USB fan pointing at it, 29c

For this build I haven't applied any BIOS power tuning, so probably consuming ~12-15w, I'd got it down to 9w on an earlier Proxmox build with the BIOS settings earlier in this thread. The Proxmox build ran a bit cooler again.

 

[pigr8]

i started fixing the x86-p5 that came in, temps were not bad but could for sure be improved.. there was a lot (i mean A LOT) of thermal paste, i removed everything and checked the cap, it's noticeable..



i would say at least .1mm, but i think it's intended to be so because there are some long spacers on the cpu itself applied not to crush the die



after removing all the paste i also sanded the case where the stock copper shim goes, it's anodized aluminium and not painted so it should not make a big difference but i sanded it down anyway





i had no kryonaut left so i applied some arctic paste and remounted everything.. i removed the cwwk sticker on the nvme drive ad underneat i found what looks like a copper thermal adhesive sheet, i left it on so it can be cooled.



idle went down to 29°C (24°C ambient) and under stress test after 10 minutes it boosts to 3.4ghz single core at max 44-45°C





i would say it's fine, it can go in production replacing the 4 port n5105.

 

[alaricljs]

Does anyone know what this is as it is the same on every similar box it appears?

Mine as another example
acpitz-acpi-0
Adapter: ACPI interface
temp1: +27.8°C (crit = +110.0°C)

Random temp sensor somewhere on the mobo.

 

[AnthonyUK]

Random temp sensor somewhere on the mobo.

It reads the same temp on every board that has been posted which is strange, no?

 

[rinze24]

1. I'm using 6 port n305 and my idle without fan was averaging around 50s-60s with some services installed (opnsense, adguard, vaultwarden, omada, nginx, uptime kuma, homepage). I repasted with arctic mx-6 but temperatures didn't change much. Adding a USB fan helped though. With fan idle temps are around 35-42. I think others here have tried adding copper shims and all to close the gap, that's something I have not tried yet.

2. I don't think you can passthrough the sensors into opnsense from the host machine. But you can try referencing this reddit post (Reddit - Dive into anything) to add temps into your pve summary page which is what I have done.

Hello, can you share your code in pvemanagerlib.js? your is more neat and not out of place to see. I cant get it right with valueField and maxField, how did you put your value in there?

 

[terencelzw]

Hello, can you share your code in pvemanagerlib.js? your is more neat and not out of place to see. I cant get it right with valueField and maxField, how did you put your value in there?

There you go sir. You can reference more from this repo as well (pve-mods/v8.0.4-pwt4.0.6/patches at main · alexleigh/pve-mods)

{
   itemId: 'cputemp',
   iconCls: 'fa fa-fw fa-thermometer-half',
   title: gettext('CPU temp'),
   valueField: 'cputemp',
   maxField: 'cputemp',
   renderer: Proxmox.Utils.render_node_temp,
},
{
   itemId: 'nvme1temp',
   iconCls: 'fa fa-fw fa-thermometer-half',
   title: gettext('NVMe1 temp'),
   valueField: 'nvme1temp',
   maxField: 'nvme1temp',
   renderer: Proxmox.Utils.render_node_temp,
}

 

[lgigante]

Hey folks,

First post here, and I am a bit lost.

I am looking at the CWWK 2-port so I still have USB 3.0.

My question is - if I use proxmox and dedicate both interfaces to Opnsense, can I still add more VMs (like a Ubuntu server) to run other stuff and still have networking on those other VMs?

Thanks!

 

[rinze24]

There you go sir. You can reference more from this repo as well (pve-mods/v8.0.4-pwt4.0.6/patches at main · alexleigh/pve-mods)

Thanks, i thought the values are just simple 0-100 or 0-1

 

[AnthonyUK]

Hey folks,

First post here, and I am a bit lost.

I am looking at the CWWK 2-port so I still have USB 3.0.

My question is - if I use proxmox and dedicate both interfaces to Opnsense, can I still add more VMs (like a Ubuntu server) to run other stuff and still have networking on those other VMs?

Thanks!

I don't think these devices support SR-IOV which would be required to share passed-through devices across VMs but it is worth checking.
The alternative is to dedicate your WAN port and share the LAN port across VMs.

 

[dunoster]

Hey folks,

First post here, and I am a bit lost.

I am looking at the CWWK 2-port so I still have USB 3.0.

My question is - if I use proxmox and dedicate both interfaces to Opnsense, can I still add more VMs (like a Ubuntu server) to run other stuff and still have networking on those other VMs?

Thanks!

I'm running mine with ESXi 8 , didn't enable passthrough for the LAN ports, virtualised opnsense runs fine together with 4 other VMs.

 

[Beanhead]

Thought this was worth a shot for a fan. 140mm USB fan

MULTIFAN S4, Quiet USB Cooling Fan, 140mm - AC Infinity

Managed to get one off ebay for £14.99.

Not the highest RPM or Airflow, but I need it to be nice and quiet as it will be in my office. Figured I can just stand the fan behind the unit to push air past it and that should be enough to bring the temps down a fair amount.

Will feedback here once I have it.

So the fan arrived. I have got the fan running on Low speed, I cannot hear it. The Fan is sitting on the shelf just in front of my unit, blowing air on to the unit.

Within a couple of minutes the case is instantly cool to touch and the system temps have dropped substantially across the CPU and the NVME.

CPU's were running at 47-50c
NVME was running at 53-55C

CPU's now running at 33-35C
NVME now running at 42-44C

The fan is caged on both sides, and has the silicon feet fitted already, so 100% designed to be used stood up like this.

Running the fan on High in this position doesn't make any further difference to the temps, just adds some noise. Still not that loud, but audible, when it wasn't audible on low. I am sitting about 3-4 feet away from it.

 

[moonman]

Within a couple of minutes the case is instantly cool to touch and the system temps have dropped substantially across the CPU and the NVME.

CPU's were running at 47-50c
NVME was running at 53-55C

All temps are already within spec, it's far from overheating. NVMe drives are usually rated to 80C, CPU can go to 100C. If the heatsink/case is warm to the touch then it's doing its job - dissipating the heat. The only reason I would consider a fan is if you live in a hot climate and temperatures go above 35C. My 5950x gets way hotter than this for hours at times (high 80C) and it's still fine years later. The only issue that I can think of with these temps is crappy ram that needs to run at lower temps in order not to produce any errors. Just buy a good ram stick and forget about it

 

[Beanhead]

All temps are already within spec, it's far from overheating. NVMe drives are usually rated to 80C, CPU can go to 100C. If the heatsink/case is warm to the touch then it's doing its job - dissipating the heat. The only reason I would consider a fan is if you live in a hot climate and temperatures go above 35C. My 5950x gets way hotter than this for hours at times (high 80C) and it's still fine years later. The only issue that I can think of with these temps is crappy ram that needs to run at lower temps in order not to produce any errors. Just buy a good ram stick and forget about it

Yeah, I know it wasn't overheating, my feedback was for other members of the group more than anything. I think it would be a good solution for anyone looking for a fairly unobtrusive way to cool their units.

Though, TBH I like to run stuff as cool as possible as in the end it will prolong life expectancy. The fan is inaudible from where I sit, so it's no bother just to leave it there running. It consumes less than 1w of power running on the low setting, which is also good.

Also, I have spare capacity to run more VM's, so who knows how the temps will perform in the future once it's running under more load.

RAM is Crucial, NVME is Crucial, so not the most expensive, but not what I would consider crap, so hopefully all good there.

Cheers.

 

[F W]

Opnsense 23.1 is finally running after a longer fight:

Setup of 23.1 failed, I got constant errors about msdos_fs not accessible after the partition creation. According to google a common "feature" of FreeBSD13. So I started with 22.7 and went along the full upgrade road to 23.1_5

If you fail to install opnsense often enough, even 22.7 reacts with file copy problems/ partistion errors. Solution is to fire up a Gnome GPartd iso and remove all existing partitions and then start from scratch.

If you encounter read/write issues/crashes during installation, use default settings for PL1/PL2 CPU and package. No idea why this happens, maybe some heat spikes in the core leading to this. Have to monitor this a bit and see if it also happens during runtime.



Bios settings (at the moment):

PCIe Configuration: Port 1: ASPM auto L1 Substate: disabled L1 low disabled
PCIe Configuration: Port 7,9: ASPM auto, L1 Substate: disabled L1 low enabled
Port 10,11,12: disabled

-> don't enable L1 Sub or L1 low on port 1, it leads to drive read errors
-> don't enable L1 Substae on port 7,9 - the NIC become invisible/unusable
-> leave port 10,11 enabled if NIC #3,#4 are needed

SATA diasbled, USB HSII on xHCI disabled, HD audio disabled, Serial I2C5 controller disabled, eMMC 5.1 Controller disabled, Sensor Hub type None

CPU Config: Intel VMX disabled, PECI enabled
GT - Power Management: Maximum GT freq 100MHz

CPU Power management: Platform PL1: disable, PL2 disable
CPU Power management -> Turbo options -> Package Power PL1 = 6000, PL2 = 20000
-> need to run some more stability tests on the PL settings. They have no influence on the idle power consumption but define the upper limits of the CPU

unplug monitor after changing primary display in opnsense to serial. the system otherwise hangs during boot. Unplug the keyboard.



Tunables in opnsense:

hw.acpi.cpu.cx_lowest = C3
hw.ibrs_disable = 1
vm.pmap.pti = 0
dev.hwpstate_intel.0.epp = 99, dev.hwpstate_intel.1.epp =99, dev.hwpstate_intel.2.epp =99, dev.hwpstate_intel.3.epp = 99

-> hw.acpi.cpu.cx_lowest can be set down to C8, however I haven't seen any relevant power reduction below C3 and it takes longer to wake up a cpu from C8 than from C3, so setting C8 might result in some data transfer delays
-> dev.cpu.3.cx_lowest changes the same flag like hw.acpi.cpu.cx_lowest, so only one needs to be set


Power consumption, 1 NIC connected, opnsense idle and not filtering anything: 7W at the all

When "PECI" is enabled, "EPB Override Over PECI" should be enabled or disabled?
For those PCIe ports of built-in Ethernet card, "L1 Low" should be enabled or disabled?
Thank you!

 

[ExplodingLemur]

I just received the i3-N305 version of the X86-P5 (Ordered September 23, shipped September 25, delivered today October 5).

I couldn't resist and tore it down right away, finding the exact same disappointing cooling solution that everyone else has already documented. I'll re-paste it with Arctic for now, but once I get my mill set back up I plan to take a skim cut of the pocket in the case to get rid of the anodizing, and get a piece of copper bar stock and face it down to just the right size to fit between the case and CPU. The copper shim this thing came with is nowhere near flat and covered in scratches and dings.

 

[pigr8]

I just received the i3-N305 version of the X86-P5 (Ordered September 23, shipped September 25, delivered today October 5).

I couldn't resist and tore it down right away, finding the exact same disappointing cooling solution that everyone else has already documented. I'll re-paste it with Arctic for now, but once I get my mill set back up I plan to take a skim cut of the pocket in the case to get rid of the anodizing, and get a piece of copper bar stock and face it down to just the right size to fit between the case and CPU. The copper shim this thing came with is nowhere near flat and covered in scratches and dings.

View attachment 32035View attachment 32036View attachment 32037View attachment 32038View attachment 32039View attachment 32040View attachment 32041View attachment 32042

yeah but once a good thermal paste is applied correctly it's actually fine, i have my unit (with the n100) already in production (impressed by now) without overheating issues, temps after stable between 33 and 44 max, but i have a fan on the back inside of the case.

i made a quick simple back cover with integrated 80x10mm silent fan connected to one of the 12v fan port, it mounts replacing the original metal cover using the same provided screws and the fan is hold in place by simple pins.. i have one of the 4 corner of the fan cutout because of the lan ports interfeering and i have it wall mounted.

the something like that



if someone need the stl/step i can provide it no problem.

 

[spongi]

Yesterday I also got the RJ42 with N305 CPU from HUNSN (Amazon Germany). During testing, I have noticed that it is overheating. If this is happening, it just freezes and I have to hard reset it. I have ordered it with the Fan Adapter Kit and will mount a Noctua 120mm Fan on top of the fins (SATA SSD installed inside), but I'm just wondering how this can happen. It's sold as fanless, but as it seems it can't be deployed fanless, at least when it's fully configured (2x NVME, 1x SATA, 32 GB DDR5 RAM). I have also seen that those units are sold with an i7 CPU, wondering how this should work

 

[AnthonyUK]

Yesterday I also got the RJ42 with N305 CPU from HUNSN (Amazon Germany). During testing, I have noticed that it is overheating. If this is happening, it just freezes and I have to hard reset it. I have ordered it with the Fan Adapter Kit and will mount a Noctua 120mm Fan on top of the fins (SATA SSD installed inside), but I'm just wondering how this can happen. It's sold as fanless, but as it seems it can't be deployed fanless, at least when it's fully configured (2x NVME, 1x SATA, 32 GB DDR5 RAM). I have also seen that those units are sold with an i7 CPU, wondering how this should work

Just read through this or any of the other 'fanless' box threads

The QC is not so great so some boxes they need some minor adjustments to ensure the CPU to heat spreader surfaces have zero space between them. Using a better and adequate amount of thermal paste is also recommended.