CWWK/Topton/... Nxxx quad NIC router

[Zorbert]

I am glad you had good luck with your crucial ram, hopefully this week I will have my new PSU and ram to test. I got one PSU recommended around page 100 or so(the anker) and I don’t know if it’s the anker block or the cord itself but the system won’t power on with it hooked up. So I ordered another one that was linked which is supposedly a. More reputable PSU maker.

I hope it isn’t the SSD, I did order another one just in case however I didn’t really want to dedicate anything with space to this box which is supposed to just be running my lan to wan for me.

 

[KevinR]

From reading many many threads, it seems the CWWK built devices (board ids CW_**), including ones sold by topton, etc appear to get on well with crucial RAM. Often explicitly recommended by CWWK themselves

Meanwhile the BKHD_ devices, mainly sold by topton and their cohort, seen to dislike crucial. These sellers often recommend Samsung or Hynex.

Topton et al also sell MW_ model devices, it's unclear who the OEM(s) are and which memory camp they are in.

Annoyingly, at least in Europe & UK, Hynex have stopped selling memory in retail form. It might still be in OEM channels or on other brands dimms.

 

[Zorbert]

From reading many many threads, it seems the CWWK built devices (board ids CW_**), including ones sold by topton, etc appear to get on well with crucial RAM. Often explicitly recommended by CWWK themselves

Meanwhile the BKHD_ devices, mainly sold by topton and their cohort, seen to dislike crucial. These sellers often recommend Samsung or Hynex.

Topton et al also sell MW_ model devices, it's unclear who the OEM(s) are and which memory camp they are in.

Annoyingly, at least in Europe & UK, Hynex have stopped selling memory in retail form. It might still be in OEM channels or on other brands dimms.

I would like to express that so far I do agree with what you have noted here and referenced. Thank you for your knowledge on this matter. I ordered both a Samsung and a Hynix ram stick(I cancelled the Samsung because it kept getting delayed in shipping.) the 8gb SK Hynix so far has lasted in my pfsense N150 box for 15 hours since I tossed it in yesterday. It isnt my record yet, but I am hoping that the results will continue without a restart.

I purchased 2 Topton n100 barebone units and the seller messaged me reach time to use oem Samsung or Hynix ram. In one unit, I used an 8 GB ddr5 5600 stick and 16 GB ddr5 Patriot 5600 made with Samsung chips in the other. Both units have been running bare metal opnsense for many months without any crashes or random reboots.

I have seen some reports of crucial not working for some people.

Audit13 I want to thank you so far with helping me troubleshoot my issues. I was going back through the forum while waiting for my ram stick to arrive, but now that the ram stick is here I was able to toss that in and restart the topton N150 running PFsense. As I stated above I am at about 15 hours uptime, I am hoping this is a cure for its random restarts/freezes. I will keep you all updated if you are interested. However it does appear that what some people were told or found out is true, that Topton needs Samsung/HK Hynix and the CWWK prefers the Crucial.

 

[TrevorH]

I did test it with a fresh default install of pfsense and it only lasted about 6 hours before it rebooted.

I see many of the most common complaints about these units seem to concern the quality of the power supply and the symptoms you describe do seem to me to be consistent with power issues. It could also be consistent with RAM problems or motherboard issues but the shipped PSUs seem not to have a great reputation.

 

[Zorbert]

I see many of the most common complaints about these units seem to concern the quality of the power supply and the symptoms you describe do seem to me to be consistent with power issues. It could also be consistent with RAM problems or motherboard issues but the shipped PSUs seem not to have a great reputation.

I do also have a new powersupply on the way, however we are at a day and a half uptime with the hk Hynix Ram. I am hopeful that is the cause and it is corrected. I am currently at a day and a half uptime.

to note for the powersupply I did get the anker one someone referenced near page 100 or so I think of the forum, however with two different cords I also purchased, the mini pc would not turn on at all while connected to it. The Anker PSU/power bank is rated for over 60watts and it was the same model as was linked in this forum from what I can tell. When that didn’t work I ordered a new full powersupply, which was also linked prior in the low 100 pages. That hasn’t arrived yet to test. Either way if the ram keeps up then I will probably not need to have wasted the money on the PSU.

All in all it cost me more than I expected due to the wasted crucial purchase and a second PSU I may not need. Which isn’t too crazy estimate 50 bucks on the PSU and 25 on the ddr5 x2 for a wasted stick. I did also get random other “mods” people mentioned to continue troubleshooting like an extra copper shim, a copper heatsink for the ram, and a coated copper heatsink for the m.2 ssd. All of which were also linked prior, we will chalk those up to 50 bucks total. However if I was super efficient I would have done these one at a time, but I am impatient and started ordering a bunch of testing materials/solutions at once to speed up trouble shooting. So we can blame impatience on the waste of money.

If the ram and new PSU don’t fix the problem I will probably just toss it in the trash and go with another solution.

 

[slybunda]

I do also have a new powersupply on the way, however we are at a day and a half uptime with the hk Hynix Ram. I am hopeful that is the cause and it is corrected. I am currently at a day and a half uptime.

to note for the powersupply I did get the anker one someone referenced near page 100 or so I think of the forum, however with two different cords I also purchased, the mini pc would not turn on at all while connected to it. The Anker PSU/power bank is rated for over 60watts and it was the same model as was linked in this forum from what I can tell. When that didn’t work I ordered a new full powersupply, which was also linked prior in the low 100 pages. That hasn’t arrived yet to test. Either way if the ram keeps up then I will probably not need to have wasted the money on the PSU.

All in all it cost me more than I expected due to the wasted crucial purchase and a second PSU I may not need. Which isn’t too crazy estimate 50 bucks on the PSU and 25 on the ddr5 x2 for a wasted stick. I did also get random other “mods” people mentioned to continue troubleshooting like an extra copper shim, a copper heatsink for the ram, and a coated copper heatsink for the m.2 ssd. All of which were also linked prior, we will chalk those up to 50 bucks total. However if I was super efficient I would have done these one at a time, but I am impatient and started ordering a bunch of testing materials/solutions at once to speed up trouble shooting. So we can blame impatience on the waste of money.

If the ram and new PSU don’t fix the problem I will probably just toss it in the trash and go with another solution.

if you dont know how to do isolated testing on these devices you would be better off getting an off the shelf solution like an edgerouter etc.

 

[Zorbert]

if you dont know how to do isolated testing on these devices you would be better off getting an off the shelf solution like an edgerouter etc.

Yeah makes sense you should have to do isolated testing and debugging for China on an all in one solution since they do not provide any documentation with the device. Guess I should test each resistor to make sure it is good? Thought my testing one component at a time was a pretty acceptable practice in troubleshooting the root cause. Which did end up determining the problem being the memory module from what I can tell, however your advice was very helpful thank you.

 

[slybunda]

how are you testing the memory? you running memtest86 from a boot usb? put screenshots here of the result

 

[Zorbert]

how are you testing the memory? you running memtest86 from a boot usb? put screenshots here of the result

That is correct I ran memtest86 from the bootable usb while in the device. I did not test the new memory yet, which is my next step. I also removed the SSD and tested it with crystal disk info in regards to its health which it is a healthy boy, that was placed back into the device to continue work.

 

[slybunda]

how long you run the memtest for? overnight?

 

[jauling]

Hey @splifingate , have you attempted any thermal/power capping on your N305? I've been waiting a long while, which might will be forever, for a 4port version with N300 or N350. Not sure if they will ever come. I'm running the N100 version at the moment, and the 8 cores in the N30x series has me super interested. I'd like to be able to run a N305 version at basically the same thermal levels of my N100 and have more performance.

Have you tried setting super low PL-values in your N305?

Yes.

I see much of the discussion around these low-power units involving Router Duties.

My 'router' is an EdgeRouter 4, and it does the job well. Uptime in is years. I'm using AT to T 1G fiber, and I have specifically configured the ER4 to entirely replace the RG (dls-ish!). I have kept the ONT as-is, as replacing it is a whole other can-of-worms.

My n305 effectively replaced a double-handful of old, bastard comps I was trying to Frankenstein back onto functionality.

 

[splifingate]

Hey @splifingate , have you attempted any thermal/power capping on your N305? I've been waiting a long while, which might will be forever, for a 4port version with N300 or N350. Not sure if they will ever come. I'm running the N100 version at the moment, and the 8 cores in the N30x series has me super interested. I'd like to be able to run a N305 version at basically the same thermal levels of my N100 and have more performance.

Have you tried setting super low PL-values in your N305?

I have not adjusted any PL's in my Unit *shrug*

Mine idles around 15W which is OK by me.

 

[mrpops2ko]

i've got the n305, it idles in proxmox with zero load at 6w (5w without a fan). doing plex transcoding sees +10w (so 15 total) and then +1 for a fan.

in a full stress test it can go as high as 40w i think, but i also don't think these units are stable enough to do constant stress test workloads or maybe my PL limiters are restrictive

 

[AlexH80]

That is correct I ran memtest86 from the bootable usb while in the device. I did not test the new memory yet, which is my next step. I also removed the SSD and tested it with crystal disk info in regards to its health which it is a healthy boy, that was placed back into the device to continue work.

Did you ever end up getting this resolved? After 9 solid months of operation, my N100 unit running proxmox randomly reboots throughout the day as well. Sometimes I stay up for 2 days, other times 4 hours. I replaced the power supply with a genuine meanwell. Waiting on new ram to try that too.

 

[splifingate]

I've been waiting a long while, which might will be forever, for a 4port version with N300 or N350. Not sure if they will ever come.

I missed the import of your comment:

My CWWK n305 AL-4L unit is a 4-port (2.5gbps) device that was introduced in c. 2024. I know that there are others, and there is no past-tense chasm to be scaled for what you want.

 

[slybunda]

i've got the n305, it idles in proxmox with zero load at 6w (5w without a fan). doing plex transcoding sees +10w (so 15 total) and then +1 for a fan.

in a full stress test it can go as high as 40w i think, but i also don't think these units are stable enough to do constant stress test workloads or maybe my PL limiters are restrictive

the n100 box i got ran stress tests overnight for many nights before i put it into production, ran linpack for over 24 hours, ran prime95 avx and blend, memtest86, core cycler and testmem5 anta absolute.
not a single error or crash.

currently its got 26 days uptime

but thats because opnsense is updated constantly, if i dont update it im sure this will run non stop unless there is a power cut.

 

[Octopuss]

Are there any newer variants of CWWK's N100-based X86 P5? Their website is a mess and slow one on top of that, and I never had the patience to rummage through their illogical mess of all products in one bag with no clear specs format.
I have V2 X86 P5 with N100 CPU, don't need more cores, but am always a sucker for something newer, more refined, maybe using less power and so on and on.

 

[Mike9474593]

Are there any newer variants of CWWK's N100-based X86 P5? Their website is a mess and slow one on top of that, and I never had the patience to rummage through their illogical mess of all products in one bag with no clear specs format.
I have V2 X86 P5 with N100 CPU, don't need more cores, but am always a sucker for something newer, more refined, maybe using less power and so on and on.

Yes, the X86-P6, including a N150/N355 CPU. I think there was also a V3 of the P5. Thinking about getting a P6, because my X86-P5 V1 is running burning hot, I hope the never revisions are a bit better.

 

[Octopuss]

Yes, the X86-P6, including a N150/N355 CPU. I think there was also a V3 of the P5. Thinking about getting a P6, because my X86-P5 V1 is running burning hot, I hope the never revisions are a bit better.

Hmm. Better than nothing, but that doesn't seem like a sound upgrade even to me Does it have anything else going for it than slightly newer but otherwise identical CPU with miniature frequency bump?

The V2 I have had a good contact with the copper plate and I think it ran just fine fanless. I mouted an 80mm fan on it later on regardless though.

 

[Mike9474593]

No, not a huge update. There is also a MAX version with ATX 12V and SFF-8654 port on the front, PCIe lanes are distributed differently. But I don't really get the sense of this version.
What might be interesting are the new versions of the bigger unit with 10G port(s): 1x AQC113C or 2x X550 or 2x x520/82500ES SFP+. I wish there was a version with x710 or the AQC113 had better support in BSD. These Intel chips are getting old.