laibsch

Hello, I am curious to hear how accurate the measurement of temperature is from the onboard sensors, particularly in my case for the Banana Pi M2+. While I do understand that the BPi M2+ suffered from poor choices for thermal management from the manufacturer both in hardware and software, I believe that thankfully Armbian has worked out the kinks on the software side, hasn't it? I am running the latest noble image, here's what I noticed. The temperature as shown by the sensors and armbianmonitor commands never really goes below 60° even on idle with ambient temperatures of around 25 to 30 degrees. My SBC already came with a small heatsink attached to the CPU. I found that value pretty high and inquired with an infrared thermometer but couldn't really find a spot on the board over 40°. Which makes me wonder if the reading is accurate and properly calibrated? My impression is that it is not. Furthermore, I noticed the board behaves quite well under load. For example, I stress-tested it with the yes command on all cores for about an hour or so and while the temperature went up to 95°, it appears the software was well-tweaked to only slightly throttle the CPU (verified with armbianmonitor -m) to not exceed this. I am not 100% sure about the thermometer readings, but I believe they were around 60° at the time, certainly nowhere near what sensors was showing. I have been testing the board for about a week now and I have to say that one time, the board did indeed shut down due to thermal overload. Interestingly, this wasn't even when the CPU was heavily loaded but there was a lot of IO wait. Temperature was high and it briefly shot up to 101° (ouch) which triggered an emergency shutdown to prevent thermal damage. Looked good to me, mostly. So, again, my question today is how trustworthy are the readings from the sensors, are they properly calibrated?

I understand this is an old question, but I thought it wouldn't hurt to add that indeed 1.7 Mbps is indeed very low. I get about 100 Mbps even over wifi (to the router, LAN from there to the BPi). CPU goes to about 80% on a single core for that ssh transfer.

OK, thank you for that information. I believe an Armbian PPA rather than the live patching would be a better approach. The project could have one just like this one. That's the Debian and Ubuntu way, packages, repos, versioning.

Hello everyone, I am an old-timer in Debian and Ubuntu and have uploaded and maintained packages there. I've come across an issue in a package that I use that affects 32-bit platforms such as armv7 but not 64-bit platforms. I've already analysed the problem and applied a fix, compiled the package locally and I'm happy. But I'd like to get the fix into armbian as well so others will not run into this fairly serious problem (for the package in question, it is essentially useless at the moment in noble). I looked through the various github repositories but was unable to see how to apply a patch to a userland package and have it published. Given that armbian systems seem to pull their packages from ports.ubuntu.com, is it that armbian does no patching of their own to packages but instead relies on Ubuntu and Debian to land their fixes? If that's the case, I've already started the process but I'd like to know if there is an armbian-specific way of fixing userland as well. Regards

It appears to me the BPi M2+ comes with up to three different Wifi chipsets according to https://docs.banana-pi.org/en/BPI-M2_Plus/BananaPi_BPI-M2_Plus; the AP6212 or the optional AP6181 and AP6335. AP6212 and the AP6212 only support 2.4GHz whereas the AP6335 is dual-band. Are you sure, you actually have hardware support for 5GHz?