Qotom Q10922H6 - Fanless N100, 2x10GbE, 4x2.5GbE
- Thread starter curl
- Start date
Notice: Page may contain affiliate links for which we may earn a small commission through services like Amazon Affiliates or Skimlinks.
I completely replaced the factory thermal pad with a 2.5 mm "Thermalright EXTREME ODYSSEY II" thermal pad, and temps on the AQC113s are now down to 47°C. So that's a good 26°C cooler than the highs seen with the original factory thermal pad (although to be fair, ambient temps were a bit warmer last week so that is also probably making a difference).
I think the factory pad was about 0.5mm too thin, and also not very conductive, and would definitely recommend replacing for anyone who buys this machine and observes similarly high temps on the network adapters.
I think the factory pad was about 0.5mm too thin, and also not very conductive, and would definitely recommend replacing for anyone who buys this machine and observes similarly high temps on the network adapters.
There's a metal shim on the underside of the lid where a thermal pad is meant to be applied. The dimensions of the shim are 1 cm wide and I think 11 cm long (but can't recall 100%). The thermal pad I bought was 85x45x2.5mm - and I cut it down to fit with about 2/3 left over.Might have to pick up some of those thermal pads as well.@curl Roughly what size did you need?
2.5mm is a pretty generous thickness, and when I screwed the motherboard back into the case/heatsink I could definitely feel the thermal pad getting smooshed a bit. But better to err on the side of too thick than too thin I think (the factory thermal pad was 2mm and didn't make great contact).
Thanks!
It didn't feel like it there was too much resistance when you screwed the motherboard back in, right?
The thermalpad being compressed a bit is probably normal, at least on other devices where I've seen them used.Yeah, it clearly wasn't making proper contact before.
It didn't take much force to get everything screwed in. The thermal pad I used was pretty soft, and I never had any concerns about damaging any components, putting strain on the board, or anything like that. I was initially a bit concerned that if I went too thick, it might interfere with the CPU heatsink making proper contact - but was able to get the 4 screws around the CPU fully tightened and CPU temps are fine. I did go around and tighten each screw a half turn at a time just to make sure everything was seated properly.
I received my N305 version of this box.
idle (proxmox launched, no vms) - 11-13W
with Windows Server 2025 VM idle - 17-18W
playing 4k video on YouTube - 26-33W
The enclosure is hot a f. You can't touch it for more than two seconds.
nvme during OS installation reached 92°C
also noticeably higher temps observed when running Windows Server 2025 (Windows 11) compared to Windows 10. at some point idling on Windows Server CPU reached 84°C, mostly it was at 68-72°C, nvme was 78-82°C.
On windows 10 it's currently idling at CPU 47°C , nvme 58°C.
Also idling with Windows 10 launched is 12-13W.
idle (proxmox launched, no vms) - 11-13W
with Windows Server 2025 VM idle - 17-18W
playing 4k video on YouTube - 26-33W
The enclosure is hot a f. You can't touch it for more than two seconds.
nvme during OS installation reached 92°C
also noticeably higher temps observed when running Windows Server 2025 (Windows 11) compared to Windows 10. at some point idling on Windows Server CPU reached 84°C, mostly it was at 68-72°C, nvme was 78-82°C.
On windows 10 it's currently idling at CPU 47°C , nvme 58°C.
Also idling with Windows 10 launched is 12-13W.
Last edited:
Yikes sounds pretty toasty. It's interesting that your idle power consumption is about the same as I've measured (I'm at about 17W with both AQC113 ports in use and one USB NVMe enclosure), but CPU temps are 40°C higher in your system. Are those power numbers measured at the wall? Also wondering what the ambient temps are where your machine is running?
What model NVMe drive are you using, and do you have a heatsink on it? I put a 1.5mm copper heatsink on mine and temps are pretty consistently around 42°C. But OpenWrt, which I am running, doesn't read/write to disk very heavily. And if the entire case is that hot, a heatsink inside the case may not make a huge difference.
Have you had a chance to check temps on the AQC113 chips at all? I'm a little scared to think what those might be in that environment.
Last edited:
Well like I said this Windows Server 2025 was making it really hot, but with Windows 10 VM it runs much cooler. And this is probably not the most typical use-case scenario for these passively cooled boxes.
I am using an SSD I've had for quite some time laying around without any use: Samsung 970 EVO Plus 1TB.
No heatsink yet although yeah we'll have to see how much it helps, I have a copper one, will try.
Ambient temp is 26.8°C
I don't know which ones are 10G and I am not using them yet but here's "sensors" command output
Idle, no VM's:
With Windows 10 idle SSD temp becomes 59-64°C and CPU package is 48°C.
Maybe these numbers are acceptable for such a passively cooled box. My main PC runs for years without a heatsink on Samsung 970 PRO 1TB and currently its temp is 55°C, 0% disk load reported in task manager.
Also OpenWRT is probably not comparable to Windows 10 in terms of resource consumption. My current Xiaomi router has 512Mb RAM and CPU which is 12 times less capable at single core than this N305. And this router runs just fine some version of OpenWRT, it has 4x2.5g ports and fake WiFi 7. So yeah maybe OpenWrt wouldn't push this N305 system to it's limits.
Also worth mentioning that your version is N100 vs mine N305 with 8 cores, so that could maybe also contribute to producing more heat.
In general I think that if someone will be planning to push these boxes to their limits they will definitely need some active cooling. If it's just a router - maybe it'll work fine with passive cooling, but the question is then why would you need such a powerfull processor? For 10G networking maybe? I don't know.
I am using an SSD I've had for quite some time laying around without any use: Samsung 970 EVO Plus 1TB.
No heatsink yet although yeah we'll have to see how much it helps, I have a copper one, will try.
Ambient temp is 26.8°C
I don't know which ones are 10G and I am not using them yet but here's "sensors" command output
Idle, no VM's:
Code:
coretemp-isa-0000
Adapter: ISA adapter
Package id 0: +47.0°C (high = +105.0°C, crit = +105.0°C)
Core 0: +43.0°C (high = +105.0°C, crit = +105.0°C)
Core 1: +43.0°C (high = +105.0°C, crit = +105.0°C)
Core 2: +43.0°C (high = +105.0°C, crit = +105.0°C)
Core 3: +43.0°C (high = +105.0°C, crit = +105.0°C)
Core 4: +47.0°C (high = +105.0°C, crit = +105.0°C)
Core 5: +47.0°C (high = +105.0°C, crit = +105.0°C)
Core 6: +47.0°C (high = +105.0°C, crit = +105.0°C)
Core 7: +47.0°C (high = +105.0°C, crit = +105.0°C)
enp1s0-pci-0100
Adapter: PCI adapter
PHY Temperature: +54.0°C
MAC Temperature: +54.0°C
acpitz-acpi-0
Adapter: ACPI interface
temp1: +27.8°C
enp2s0-pci-0200
Adapter: PCI adapter
PHY Temperature: +56.0°C
MAC Temperature: +56.0°C
nvme-pci-0300
Adapter: PCI adapter
Composite: +57.9°C (low = -273.1°C, high = +84.8°C)
(crit = +84.8°C)
Sensor 1: +57.9°C (low = -273.1°C, high = +65261.8°C)
Sensor 2: +59.9°C (low = -273.1°C, high = +65261.8°C)Maybe these numbers are acceptable for such a passively cooled box. My main PC runs for years without a heatsink on Samsung 970 PRO 1TB and currently its temp is 55°C, 0% disk load reported in task manager.
Also OpenWRT is probably not comparable to Windows 10 in terms of resource consumption. My current Xiaomi router has 512Mb RAM and CPU which is 12 times less capable at single core than this N305. And this router runs just fine some version of OpenWRT, it has 4x2.5g ports and fake WiFi 7. So yeah maybe OpenWrt wouldn't push this N305 system to it's limits.
Also worth mentioning that your version is N100 vs mine N305 with 8 cores, so that could maybe also contribute to producing more heat.
In general I think that if someone will be planning to push these boxes to their limits they will definitely need some active cooling. If it's just a router - maybe it'll work fine with passive cooling, but the question is then why would you need such a powerfull processor? For 10G networking maybe? I don't know.
Last edited:
enp1s0-pci-0100 and enp2s0-pci-0200 would be the two 10GbE interfaces. The 2.5 GbE interfaces don't report temperatures. Looks fine for now, but it may be worth keeping an eye on them if/when you do start using them.
I think running a VPN or IPS at 10Gbps could probably benefit from an N305. But there's a lot you can do with OpenWrt, including hosting any service you want in Docker, so lots of potential uses. For simple 10Gbs routing, I have found the N100 to be more than enough.
Yeah I noticed in the iKoolCore R2 Max review, which is pretty similar to this Qotom box in terms of hardware and design, it looks like the N305 version comes with active cooling while the N100 is passively cooled. So it sounds like the N305 can be a bit of a challenge to passively cool in a Mini PC form factor.
I'm glad to hear that the Windows 10 VM is running so much cooler. I wonder why Windows Server 2025 was so much hotter...
Last edited:
I don't know what to say, but I noticed there are some processes in the background that sometimes start to pull some resources. Like "System" or some "Packet Manager". So CPU sometimes spikes up to 50% load, disk to 40-60%. Temps fluctuate a lot, but I can say that when I left it idling for hours I saw CPU temp at some point was 84°C. Maybe it went down afterwards.
I launched AIDA64 system stability test on Windows Server, after 32 minutes temps are like this:
Code:
root@proxmox:~# sensors
coretemp-isa-0000
Adapter: ISA adapter
Package id 0: +96.0°C (high = +105.0°C, crit = +105.0°C)
Core 0: +82.0°C (high = +105.0°C, crit = +105.0°C)
Core 1: +82.0°C (high = +105.0°C, crit = +105.0°C)
Core 2: +82.0°C (high = +105.0°C, crit = +105.0°C)
Core 3: +82.0°C (high = +105.0°C, crit = +105.0°C)
Core 4: +96.0°C (high = +105.0°C, crit = +105.0°C)
Core 5: +96.0°C (high = +105.0°C, crit = +105.0°C)
Core 6: +96.0°C (high = +105.0°C, crit = +105.0°C)
Core 7: +96.0°C (high = +105.0°C, crit = +105.0°C)
enp1s0-pci-0100
Adapter: PCI adapter
PHY Temperature: +79.0°C
MAC Temperature: +79.0°C
acpitz-acpi-0
Adapter: ACPI interface
temp1: +27.8°C
enp2s0-pci-0200
Adapter: PCI adapter
PHY Temperature: +81.0°C
MAC Temperature: +81.0°C
nvme-pci-0300
Adapter: PCI adapter
Composite: +92.8°C (low = -273.1°C, high = +84.8°C) ALARM
(crit = +84.8°C)
Sensor 1: +92.8°C (low = -273.1°C, high = +65261.8°C)
Sensor 2: +100.8°C (low = -273.1°C, high = +65261.8°C)I don't think it can cool down N100 system under heavy load very well either. Typically N100 mini-PCs always have active cooling. I think these boxes could probably run firewall/routing tasks without an active cooling (will have to test that though), but definitely not staying under full load for hours.
For iKoolcore N100/N305, notice this product. You see they mention:
Fanless is N100 and that product exists for a reason.
I think N305 CPU is definitely useless without propper cooling here, you won't be able to use its full potential. I think I can try mounting some external fan on top of the box. For N100 I suspect it's the same picture in general as pretty much all the regular N100 mini-PCs on the market have active cooling. But if it was supposed to only be used as firewall/router - choosing N100 at least makes a bit more sense. Although it looks like you don't need N100 power for such tasks either.
There's no need for speculation here, since I have an N100 system and can run tests. Here are the results from running stress-ng:
Code:
root@OpenWrt:~# stress-ng --matrix 0 --tz -t 30m
stress-ng: info: [7112] setting to a 30 mins run per stressor
stress-ng: info: [7112] dispatching hogs: 4 matrix
stress-ng: info: [7112] matrix:
stress-ng: info: [7112] acpitz 27.80 C (300.95 K)
stress-ng: info: [7112] x86_pkg_temp 70.75 C (343.90 K)
stress-ng: info: [7112] skipped: 0
stress-ng: info: [7112] passed: 4: matrix (4)
stress-ng: info: [7112] failed: 0
stress-ng: info: [7112] metrics untrustworthy: 0
stress-ng: info: [7112] successful run completed in 30 mins
Code:
eth5-pci-0200
Adapter: PCI adapter
PHY Temperature: +52.0°C
MAC Temperature: +52.0°C
coretemp-isa-0000
Adapter: ISA adapter
Package id 0: +70.0°C (high = +105.0°C, crit = +105.0°C)
Core 0: +70.0°C (high = +105.0°C, crit = +105.0°C)
Core 1: +71.0°C (high = +105.0°C, crit = +105.0°C)
Core 2: +71.0°C (high = +105.0°C, crit = +105.0°C)
Core 3: +71.0°C (high = +105.0°C, crit = +105.0°C)
acpitz-acpi-0
Adapter: ACPI interface
temp1: +27.8°C
eth4-pci-0100
Adapter: PCI adapter
PHY Temperature: +51.0°C
MAC Temperature: +51.0°C
nvme-pci-0300
Adapter: PCI adapter
Composite: +43.9°C (low = -273.1°C, high = +89.8°C)
(crit = +94.8°C)
Sensor 1: +43.9°C (low = -273.1°C, high = +65261.8°C)
Sensor 2: +43.9°C (low = -273.1°C, high = +65261.8°C)
Code:
eth5-pci-0200
Adapter: PCI adapter
PHY Temperature: +53.0°C
MAC Temperature: +53.0°C
coretemp-isa-0000
Adapter: ISA adapter
Package id 0: +52.0°C (high = +105.0°C, crit = +105.0°C)
Core 0: +52.0°C (high = +105.0°C, crit = +105.0°C)
Core 1: +52.0°C (high = +105.0°C, crit = +105.0°C)
Core 2: +52.0°C (high = +105.0°C, crit = +105.0°C)
Core 3: +52.0°C (high = +105.0°C, crit = +105.0°C)
acpitz-acpi-0
Adapter: ACPI interface
temp1: +27.8°C
eth4-pci-0100
Adapter: PCI adapter
PHY Temperature: +51.0°C
MAC Temperature: +51.0°C
nvme-pci-0300
Adapter: PCI adapter
Composite: +43.9°C (low = -273.1°C, high = +89.8°C)
(crit = +94.8°C)
Sensor 1: +43.9°C (low = -273.1°C, high = +65261.8°C)
Sensor 2: +43.9°C (low = -273.1°C, high = +65261.8°C)I would say that temps were pretty well controlled under load, reaching a maximum of 71.0°C on one of the cores, and quickly dropped back to 52.0°C within 1 minute of the test ending.
I have some graphs showing temps and cpu load over time, but unfortunately I'm getting an error now when trying to upload images on this forum. What I can observe is that temps shot up in the first minute of the test to 60.0°C, and then over the next 20 minutes slowly climbed up to 71.0°C where they remained for the last 8 minutes of the test. Possibly there is some throttling occurring here to maintain temps.
Note that TDP on the N305 is 2.5x that of the N100 (15W vs 6W). That's a significant difference, and I would hesitate to jump to conclusions about the thermals of the N100 system based on experience with the N305 system.
I think that this case+heatsink is able to cool an N100 just fine for most use cases. But if your use case involves frequent, sustained periods of 100% CPU usage and/or operating in an area with very high ambient temperatures, then you may want to consider something with active cooling.
Last edited:
@curl Are your NVMe temps identical during the stress tests and afterwards? Wow, something must be wrong with my NVMe. My results are basically 18°C higher than yours, but the ambient temperature in my room is 25°C, so I'm not sure if we should subtract 6°C and consider it a 12°C difference. I also haven't repasted anything yet. Given that and the odd NVMe behavior, the difference could be less than 10°C, which is probably expected. I'll continue testing after repasting and with a different NVMe.
But if higher TDP implies more power (and therefore more heat), it should mostly happen only when that extra power is used. During tasks that consume the same amount of resources on both CPUs, the heat generated can be the same. If that’s the case, both the N100 and the N305 should have similar thermal performance for basic routing tasks, with the N305 simply offering more power if needed. I'll check whether that holds true after repasting and installing a new NVMe.
Nonetheless, I don’t think this enclosure is very good for cooling either the N100 or the N305. I believe this is true for all similar boxes. My reasoning is that typical N100 mini PCs have active cooling, and I’ve seen many reports of extremely high temps when people use an “N100 mini as a router.” I don’t think this box is fundamentally different. It might be fine for some tasks, but what I’m saying is that this enclosure can’t handle the N100’s full potential under sustained load. If you don’t need that performance, it might still work fine as a router. Then again, do we really need an N100 just for routing? Anyway, I don’t want to repeat myself—I’m glad if it suits your needs.
For me, since I actually want that extra power, I’m going to try to improve the thermals as mentioned, and then mount a fan if necessary.
Right now, based on my experience, if I were choosing something specifically for routing/firewall with passive cooling, I would look at other CPU options.
I’m currently trying to attach a 12V fan from my old PSU to this box. The 4-pin connector fits the “SYS_FAN” header, but the wiring seems different, and it looks like “SYS_FAN” only provides 5V. I did get the fan to spin by swapping some wires, but it wouldn’t report RPM in the BIOS and just ran at a constant speed. I also couldn’t find any useful BIOS settings to manage temperatures and adjust speeds. There’s a larger “SATA Power Connector” that provides 12V, but that yielded similar results. I also asked the supplier for a PDF manual, but they don't have it.
So at this point, I’m not sure if there’s a fan-type mismatch, if the board doesn’t support PWM (even though it has settings for it in the BIOS and a 4-pin header), or if something else in the BIOS needs to be enabled. I’ll try looking for information from people who have attached fans to Qotom boxes, and ask the supplier which fan I should buy.
It would be interesting to see results from a one-hour test. I notice that my temperature keeps climbing during the test, and I'm not sure where it would stop.
Well, I'm not sure what that TDP value means in real life. If we're talking about power consumption, the real numbers I saw show actual usage is quite similar between the N100 and the N305. Even some older “T” processors sometimes outperform the N100 in that regard despite having a 35W TDP.
But if higher TDP implies more power (and therefore more heat), it should mostly happen only when that extra power is used. During tasks that consume the same amount of resources on both CPUs, the heat generated can be the same. If that’s the case, both the N100 and the N305 should have similar thermal performance for basic routing tasks, with the N305 simply offering more power if needed. I'll check whether that holds true after repasting and installing a new NVMe.
Nonetheless, I don’t think this enclosure is very good for cooling either the N100 or the N305. I believe this is true for all similar boxes. My reasoning is that typical N100 mini PCs have active cooling, and I’ve seen many reports of extremely high temps when people use an “N100 mini as a router.” I don’t think this box is fundamentally different. It might be fine for some tasks, but what I’m saying is that this enclosure can’t handle the N100’s full potential under sustained load. If you don’t need that performance, it might still work fine as a router. Then again, do we really need an N100 just for routing? Anyway, I don’t want to repeat myself—I’m glad if it suits your needs.
For me, since I actually want that extra power, I’m going to try to improve the thermals as mentioned, and then mount a fan if necessary.
Right now, based on my experience, if I were choosing something specifically for routing/firewall with passive cooling, I would look at other CPU options.
I’m currently trying to attach a 12V fan from my old PSU to this box. The 4-pin connector fits the “SYS_FAN” header, but the wiring seems different, and it looks like “SYS_FAN” only provides 5V. I did get the fan to spin by swapping some wires, but it wouldn’t report RPM in the BIOS and just ran at a constant speed. I also couldn’t find any useful BIOS settings to manage temperatures and adjust speeds. There’s a larger “SATA Power Connector” that provides 12V, but that yielded similar results. I also asked the supplier for a PDF manual, but they don't have it.
So at this point, I’m not sure if there’s a fan-type mismatch, if the board doesn’t support PWM (even though it has settings for it in the BIOS and a 4-pin header), or if something else in the BIOS needs to be enabled. I’ll try looking for information from people who have attached fans to Qotom boxes, and ask the supplier which fan I should buy.
The stress test that I ran only stresses the CPU and RAM - I would prefer not to subject my bargain basement NVMe with a finite number of write cycles to a stress test. So I would not expect NVMe temps to increase except if air temps in that part of the case increased significantly, which they did not.
I believe the AIDA64 system stability test that you ran also stresses the NVMe, so I would expect NVMe temperatures to rise significantly during that test.
The Samsung 970 PRO 1TB is an older drive model, and uses a DRAM cache, which supposedly uses more energy than DRAM-less. It's possible that switching to something DRAM-less and with a more modern controller might run cooler. But if you haven't tried it yet I would first recommend putting a heatsink on it and seeing if that makes a difference. Here's a youtube video with some good advice for effectively cooling a Samsung SSD. Or you could just put on a nice, thick thermal pad and smoosh it a bit to ensure that good contact is made between all the chips and the heatsink.
Is those NIC temperatures when having the NIC plugged in or without it plugged in?
Without a cable plugged in, my Aquantia NICs idle at 50°C and 48°C respectively without any thermalpad modification/replacement. With a cable plugged in, it jumps to 57-58°C.
Basically, I'm trying to determine if it's still worth buying and replacing the thermalpad. If your reported temperature was from a disconnected state, then it probably isn't but if not, it might still be worth doing.
That's with both Aquantia NICs plugged in.Is those NIC temperatures when having the NIC plugged in or without it plugged in?
For context - I just had a look at temperatures here for the last week as the weather has gotten warmer since my earlier post (outdoor temps hit 18°C yesterday, but no idea what the indoor temps were). The highest temperature recorded on the Aquantia NIC was 52°C, and the low was 46°C. Again, this is with the 10Gbps NICs plugged in.
Looks like your unit may not be as bad as mine, where temps were getting up near 70°C. I think you would probably see some difference if you replaced the thermal pad, but may not be necessary.
I meant to ask - did you end up going with a Meanwell AC adapter instead of the one that Qotom ships? If so - has it improved system power consumption at all?
Last edited:
Ok, so that's still quite a bit lower than my readings then. Might have to order that thermal pad.
Yeah, my unit seems to have better thermal pad contact based on the temperatures and is probably more in line with what they designed it for. Still, it might be worth lowering the temperature further in the interest of longevity. It's kind of hard to judge your numbers without knowing the indoor temperature though.
Your previous post and the fact that you mentioned the outdoor temperature again in this post makes me thing that your unit is placed in a garage or something like that. Is that true?
If so, the difference I would see by changing the thermal pad might not be that big after all.I did order the Meanwell AC adapter but I haven't tried it yet. That's partially to do with the fact that I haven't migrated over to this Qotom box yet, and partially due to the fact that I didn't check the specs closely enough before ordering it to notice that the connector on it is of the 5.5 mm x 2.1 mm variety, not 5.5 mm x 2.5 mm. I have an adapter on the way that will hopefully make it fit. I didn't order the separate power supply primarily to improve power consumption (I have no tool to measure it anyway), but mainly because I thought the Dajing power supply might be of questionable quality. My initial impression of it is that it might actually be decent though, even if I haven't gotten any numbers or data to back that up.
It's in a small room that is insulated, but doesn't have heating/air conditioning so temperatures do fluctuate quite a bit. I mentioned the outdoor temps because that was the only data point I had, but realize that isn't really very useful, so I measured the ambient temperature in the room just now.
The room temp is 18°C, and the two Aquantia NICs are both at 52°C. So 34 degrees above ambient temperatures, and that's with both NICs connected at 10Gbps link speed, and light network usage (streaming some TV).
Hopefully that is a bit more useful in helping you decide whether you want to replace the thermal pad.
Ah, I see. That makes sense.It's in a small room that is insulated, but doesn't have heating/air conditioning so temperatures do fluctuate quite a bit. I mentioned the outdoor temps because that was the only data point I had, but realize that isn't really very useful, so I measured the ambient temperature in the room just now.
The room temp is 18°C, and the two Aquantia NICs are both at 52°C. So 34 degrees above ambient temperatures, and that's with both NICs connected at 10Gbps link speed, and light network usage (streaming some TV).
Hopefully that is a bit more useful in helping you decide whether you want to replace the thermal pad.
That's indeed a lot more useful. Thanks for taking the time to check!
The NIC in mine is currently at 58°C in what I believe is roughly a 20-22°C ambient temperature room. That's with just one of the Aquantia NICs connected (at 2.5 Gbps currently) and the other one disabled in the BIOS (with the assumption that disabling it saves power).
Basically, it seems like your temperatures are probably slightly better. I guess I'll have to see how much of a difference it makes to the NIC temperature to connect it at 10 Gbps. I imagine that doing so might bump up temperatures a few degrees at the very least.