UDOO Thermal Testing Results Conditions * Ambient Air temperature 22C / 73.2F * UDOO placed on oak desk surface, IR temp of surface 21c * Between the UDOO board and the desk is one of the pieces of foam that came in the UDOO packaging. Since the desk surface is notably lower than the board and the core, this would allow the desk to act as a heatsink and throw off the effectiveness of the heatsinks ability to dissipate heat on its own. UDOO Setup * The test board is running the UDOO provided Linaro 1.2 image with the UDOO provided update. * The kernel has also been patched to resolve the CPU load issues described in this thread viewtopic.php?f=2&t=149. An initial review showed this caused a couple degree drop in CPU core temp so the thermal imaging tests have all been redone. The table contains the updated data and new images have been added to the G+ Photo Album * Load is simulated by running each core to 100% using the following command Code: fulload() { dd if=/dev/zero of=/dev/null & }; fulload; read; killall dd The command is repeated for each number of cores I want to load, here is the command twice to load 2 cores Code: fulload() { dd if=/dev/zero of=/dev/null | dd if=/dev/zero of=/dev/null & }; fulload; read; killall dd * Process(s) were allowed to run for 5 minutes, then the measurements and images were taken * CPU Core temperatures are being taken from /sys/devices/virtual/thermal/thermal_zone0/temp file Based on my original testing I found 2 oddities 1.) The SAM3X8 chip got hotter than expected as I was not intentionally loading it. I plan to do further testing and possibly add a heatsink when I can get a good testing sketches together for it. (Submissions welcome) 2.) The temp of the WiFi adapter got 'much' hotter than I expected with no significant network load. As a result I added a heatsink. The table shows the temps before and after the heatsink. Table of results http://goo.gl/ut4aSa Images from the testing http://goo.gl/GQG2yg Updates: * Added heatsink to WiFi Module - before and after thermal images posted in thread and added to photo library linked above. * Added i.MX6 Core temps to table linked above, Temp readings of the actual core reported by /sys/devices/virtual/thermal/thermal_zone0/temp * Updated table and Thermal images after apply the kernel patch noted in thread: viewtopic.php?f=2&t=149
wifi temp is interesting. hmm, might look into mounting a small fan or something for my installation. 50 Deg C / 120 F is a bit warmer than I expected.
http://www.rhworkshop.com/products/ir-blue-dm Just a device that marries smartphone to Melexis MLX90620 Sensor. And yea i was shocked by the wifi temp.. When i first saw it in the thermal image i checked by feeling it and it was in fact quite hot. Maybe its just mine, does yours feel as hot when its actually enabled and connected?
cute little device. Be groovy to get one of those cheap Laser Thermometers (for parts), hook it up to some steppers and hack into it to get the values back. IE sweep it back and forth and build up data points and overlay on top of a standard visible image. Or start from scratch with a sensor or similar.
I added a heatsink to the Wifi module and it seems to be rather effective. I removed the white sticker that was on top, and placed on the bottom of the module so i didn't lose it. The top is smooth solid aluminium so it made attaching the heatsink simple. Doing so dropped the running temp of the module by 15-20F or so. I posted a before and after photo in my album above but here it is specifically. Before the heatsink i was measuring the WiFi adapter from 110-120F After the adapter i have yet to see it go over 100F i am considering adding one one to the SAM3X chip also cause its a hot spot too but i want to do more testing before i do that.
Good resource i will be doing some thermal testing for my case design, There is some useful data here.
Also noting that i have added i.MX6 Core temps to table linked above Re did the load tests for idle, then loading each of the 4 cores and noted temps. kintera, thanks, let me know if there is any other more specific parts or areas you want to see details on or any load types/profiles.
Mainly i needed baselines for none case which i have right here. though i will be using a infrared thermal probe to measure. but this shows me hot spots and where i need to focus air to.
Was there any particular program you had running to do the test? I have access to a proper thermal imaging camera at work. Would be happy to try it out for the benefit of the community. I should have a heatsink to put on the wifi module as well to see the difference.
With the new patch that was put together to address the kernel problem that was causing high "load" values even though there was no CPU usage, i did some basic tests to see if the core temp was lower than i had previously documented and it was by a couple degrees. It wasn't a lot but it was enough to where i thought i should re-do the testing. As such i have updated the original post with more details (@Ralphle79, should answer your questions). I have updated the GDoc Spreadsheet and also the G+ Album with the relevant images. i deleted the old images leaving only the ones that match up with the data in the table. I left the various images for the wifi adapter too. Please feel free to pass along any other info, or suggestions. I do want to put the SAM3X8 chip to work, so if anyone could help me with a sketch that should push that chip to its limits let me know. I wanted to also do some overclocking tests, but i ended up spending the day on redoing this. I am off work tomorrow, so if i can actually get the overclocking to work, i will proceed with that testing.
I did the same test, but had all 4 cores running for 5 mins. Plus I tried downloading a 70mb file at the same time to get the wifi a good run. I wanted to get this image the same funky colours as the next, but the stupid software moves the markers around and seems to be giving the wrong temps when I move them back and export it back out. I stuck a heat sink on the wifi module - the little "M" under it is a bit of paper to hold it from sliding off. Had it on an angle so the IR Camera could see it straight on. Also, took a photo of the Sata drive attached to the underside of the MDF board.
FLIR... sexy. Did you record the ambient and on-die temps during that run? Very useful info indeed, thanks to Draco & Ralpie for their work so far.
great comparison, however im worried i wonder how far off mine are. In other tests i had done my imager wasnt that far off.. i expect a variance of a few degrees, but your images are showing 20C / 36F degrees hotter than mine. That is quite a difference
I wasn't sure how to record the on-die temps. I'm not familiar with linux at all. This is my first real play with it. If someone could tell me how to do it, I'll give it another run and see if there was much difference. Oh, and I forgot to mention those images were before the kernel patch was added. If I have some spare time i'll try it again - logging it and now that I have the new patch added. I didn't have time to do the patch before the test last week.
Its all in my original post, what i did and how feel free to take follow the same route and see where it leads
Draco would you like to share a comparison table with the old results and the latest results after applying the Kernel patch? I'm trying to follow you guys, not easy I must say!
I was able to pull a historical version (Thank you google docs) if you see the spreadsheet now you will see the old data on the bottom, you will notice a distinctly lower temp at lower uses.
