While it's for a different mini PC, the post linked below shows that sometimes a drained CMOS battery can cause weird issues. Might be worth trying if you haven't already tried it.
Qotom Denverton fanless system with 4 SFP+
- Thread starter is39
- Start date
Notice: Page may contain affiliate links for which we may earn a small commission through services like Amazon Affiliates or Skimlinks.
While it's for a different mini PC, the post linked below shows that sometimes a drained CMOS battery can cause weird issues. Might be worth trying if you haven't already tried it.
I think you might be on to something. No idea why this caused it to turn off, though.
Now to find out where to source a Newsun CR2032-55-2 with either a tiny Molex or JST connector and hope I get the connector type and polarity right. Why couldn't they just have used a normal cell or holder?
Yeah, it's definitely a bit weird.
It sounded like it might POST with the battery disconnected too, at least that other mini PC. Might be worth trying before the new battery arrives.
Do you mean that the cell is different from a regular CR2032?
Those CR2032 covered in soft plastic with a cable and connector on it are becoming more common on modern consumer PC motherboards now too due to space contraints, so they hopefully shouldn't be that hard to find.
Does not boot with it disconnected. SFP+ bay lights flash and there's at least one red light on the board flashing.
Sorry for the crap photos.
I have taken a video and sent it to Qotom.
Thanks for your help, blunden.
Ah yes, those are the type I tried to describe. Should be fairly standardized as far as I know, not some proprietary nonsense. Like I said, they are starting to become somewhat common on motherboards from well known brands too. My MSI X670E Carbon AM5 motherboard has one for instance.
I would guess it's simply a regular CR2032 with wires either soldered to it, or simply taped to the battery. If you can't source a new one, you could probably just disassemble it, replace the CR2032 and tape it back up.
I just found this thread as my Qotom Q20331G9 also died last week. First it turned itself off but would still boot again, then it went completely dead. I suspected the power supply, but the supply measures fine and I also tried using a lab power supply, and it clearly doesn't even try to boot (12 V current draw is something like 30 mA).
The battery on mine measures 3 V, and trying to boot with it disconnected also does nothing. I even tried supplying 3 V from another lab power supply to the battery input, but no luck.
The battery on mine measures 3 V, and trying to boot with it disconnected also does nothing. I even tried supplying 3 V from another lab power supply to the battery input, but no luck.
Does anyone have experience with the unit running the C3808? I would like it to handle routing + firewall + wireguardVPN + an attached DAS like Patrick's review as a NAS setup using proxmox virtualisation. While the Atom processor is no Xeon, I would imagine having the 12 core @2.0Ghz and 64Gb of RAM would mirror the Netgate 8300 BASE setup? I understand many here like dedicated boxes but I am trying to find a start point before splitting things into separate boxes/redundancies. I have a symmetrical 10Gbps provider. Was hoping to hit 7-8Gbps with the Qotom minipc.
I have no experience with the C3808 (but others in here do), but I manage roughly 10 Gbps in a single direction with the C3758 (8 core, 2.2 GHz) and roughly 16 Gbps bidirectionally in internet speed tests on bare metal. I haven't tested in a controlled environment (LAN) though, and neither have I tested virtualized performance.
I have no idea how it compares to the Netgate appliance you mentioned though.
I run Wireguard on mine too, but I have yet to have the opportunity to speed test that in any meaningful way since it's already deployed as my main router.
Those results were in VyOS with software flow offloading btw. I would expect BSD based OS:es to be slower, based on a lot of results from /r/Init7.
An update:
Replacing the CR2032 did not help, still not booting and SFP+ & motherboard lights blinking some kind of error code.
Qotom agreed on an RMA but the board has been held by Chinese customs for a few days now.
Replacing the CR2032 did not help, still not booting and SFP+ & motherboard lights blinking some kind of error code.
Qotom agreed on an RMA but the board has been held by Chinese customs for a few days now.
got the model with Atom C3558, 2 SFP 1G, 2 SFP+ 10G.
It gets very hot to the touch being idle, reports shows 53-55 degrees.
Comes with a small fan on the case bottom but doesn't seem effective.
Wish they stuck 2 80mm intake and exaust fans on the bottom, would have helped a ton.
Is there a SSD 2.5 mount slot somewhere I'm missing? I couldn't find the mount holes that line up to the SSD screws
It gets very hot to the touch being idle, reports shows 53-55 degrees.
Comes with a small fan on the case bottom but doesn't seem effective.
Wish they stuck 2 80mm intake and exaust fans on the bottom, would have helped a ton.
Is there a SSD 2.5 mount slot somewhere I'm missing? I couldn't find the mount holes that line up to the SSD screws
It's supposed to be screwed to the bottom cover, at least for those of us with the case without the bottom fan. What if you rotate your SSD 90 degrees from your current orientation and try the 4 screw holes roughly where it is in the picture? A bit too wide perhaps?
rotating 90 degrees the holes are too wide yeah. I think those grommets are for a 3.5 inch HDD drive
I see. That's an unexpected change. My unit has screw holes for a 2.5 inch SSD, which is also what I'm using in it.
Is it a good idea to get thermal pads on top then plop a big heatsink on there? its toasty to the touch, for now I've put an ice pack on there which lowers it by 15 degrees
Oh right, you got the new chassi. That might explain the different hole pattern.
Is the temperature really a concern though? It's an embedded CPU designed to handle high temperatures after all.
yeah added dual 80mm fans, running at 16 degrees lower much better
it was hitting 55 degrees without the 10G SPF+ ports being used, I'd imagine the NIC would fail or thorttle soon after using itOh right, you got the new chassi. That might explain the different hole pattern.
Is the temperature really a concern though? It's an embedded CPU designed to handle high temperatures after all.
I see. My C3758 version idles at 45-46°C. Yours is clearly a bit hotter though, so perhaps a repaste could help?
I use it as a 10 Gbps router just fine without extra cooling. The X553 NICs are integrated in the SoC as far as I know, so I imagine it's fine.
I just wonder, since this have long past and trouble have probably been resolved. But I wonder, when people says taking the CMOS battery out didn't fix it. did they account factor in the memory training also? I means, if CMOS battery is out then the BIOS wouldn't have any saved data thus wouldn't that means memory training is required every time after power off also. Could it be that they thought the board is dead because it won't come back up right away and didn't wait for memory training to complete?
I just recently noticed something that I find very weird. I did some speed tests and noticed that my upload was suddenly only 4-5 Gbps, while my download remained steady at 9.4-9.6 Gbps.
Testing with other local devices, I get full speed between them. However, testing from the router to those same local devices show the same speed degradation on the upload. Upload stays at a 4-6 Gbps, no matter the amount of streams. The suggests to me that testing from the router like that, which I know isn't really a measure of routing performance (but I wanted to try to remove variables), might make me run into a CPU limit.
Are people seeing similar performance limits when the Qotom box is forced to generate the packets during an iperf3 test between the Qotom and another 10 Gbps connected client?
Seeing as I know that I had no problems getting it to route at full speed when I originally installed it, I figured it might be cause by a performance regression in VyOS. However, reverting to the old image and configuration I initially used back then, it performed even worse (in both directions) for some reason.
Outlier: I can get iperf3 upload results to an internet facing server relatively close by upwards of 8.7 Gbps through the router if I use 10-12 streams, but I see quite a lot of retries. Seems like those could be an issue?
Testing with other local devices, I get full speed between them. However, testing from the router to those same local devices show the same speed degradation on the upload. Upload stays at a 4-6 Gbps, no matter the amount of streams. The suggests to me that testing from the router like that, which I know isn't really a measure of routing performance (but I wanted to try to remove variables
), might make me run into a CPU limit.Are people seeing similar performance limits when the Qotom box is forced to generate the packets during an iperf3 test between the Qotom and another 10 Gbps connected client?
Seeing as I know that I had no problems getting it to route at full speed when I originally installed it, I figured it might be cause by a performance regression in VyOS. However, reverting to the old image and configuration I initially used back then, it performed even worse (in both directions) for some reason.
Outlier: I can get iperf3 upload results to an internet facing server relatively close by upwards of 8.7 Gbps through the router if I use 10-12 streams, but I see quite a lot of retries. Seems like those could be an issue?