03-23-2016, 01:36 PM
Got mine the other day. The board is bent, but it still boots RemixOS the gigabit ethernet is working. I don't have a class10 sd card yet, so I haven't done any performance testing.
Got your PINE64 board yet? Post here!
|
03-23-2016, 01:36 PM
Got mine the other day. The board is bent, but it still boots RemixOS the gigabit ethernet is working. I don't have a class10 sd card yet, so I haven't done any performance testing.
I have Pine64 developers edition. I install radiators...
Отправлено с моего SM-T325 через Tapatalk 62 C degrees temperature on SOC when viewing full HD movie. I note as I believe it is necessary, I do not use thermal glue. If you use thermal glue, remember that its thermal conductivity is worse than that of the thermal paste. If you want I can post instructions on how to stick to the radiators Pine64 with using the thermal paste and super glue. But it should be done carefully, and take good radiators (preferably copper) as they remove the problematic back))) Отправлено с моего SM-T325 через Tapatalk
03-29-2016, 06:23 AM
(03-27-2016, 11:11 AM)Artyom Wrote: I have Pine64 developers edition. I install radiators... Hi Artyom, Why didn't you put a heatsink on the A64 ARM CPU? Copper or Alu heatsinks ? The colour also have en impact on the thermal resistance. Black is better to radiate heat than blank surfaces. The voltage regulator next to the micro USB connector might also benefit a added heatsink. (only a suggestion. I'm a April backer and haven't got my Pine A64 jet)
03-29-2016, 06:32 AM
the Thermopad between the processor and heatsink derives maximum 1/10 of the heat as the heat sink. Since the color is not important. The effect is equal to zero.
03-29-2016, 06:43 AM
Nobody has done much real testing on types of heatsink materials and coatings, but i assume the more surface area you have, the more you can dissipate. Color or anodizing wouldn't do much, but having a rough surface would likely provide very slightly an increase in heat dissipation, but unless you're trying to stress test your Pine64 i wouldn't fret over little things like that. Just get some bare aluminum heatsinks, and maybe a fan if you want to keep it cooler.
If I've helped you with something, please leave a rating for my responses.
03-29-2016, 07:01 AM
(03-29-2016, 06:23 AM)gbjensen Wrote:(03-27-2016, 11:11 AM)Artyom Wrote: I have Pine64 developers edition. I install radiators... This copper radiators from my old video card Asus Riva TNT2. I just tried to run and load Pine64 under Android. As it turned out, he averaged more heats SOC and RAM than linux without much stress. Then simply play FullHD video. And I realized that it is necessary cooling. Since according to measurements from the Android software that comes SOC RAM and heated very well. After that I found my old video card and took off her radiator. And then put on the SOC and RAM modules. I decided to use the passive cooling. If you need tests, describe in detail the methodology of testing and I'll do. (03-27-2016, 11:11 AM)Artyom Wrote: 62 C degrees temperature on SOC when viewing full HD movie Well, the most important information is missing since jemk's commit to add HW accelerated video decoding is already a few weeks old: https://github.com/linux-sunxi/libvdpau-...e85f2c97c7 If you already made use of Cedrus acceleration in Linux (or used Android where this was HW accelerated all the time) then 62°C is freaking hot -- haven't measured it with A64 but with H3 temperature between idle and playing jellyfish-90-mbps-hd-hevc.mkv (324MB in size, 90 Mbps bitrate, 1920x1080, HEVC, Main profile, Level 5.0, Tier: high) just differs by a few °C and also power consumption just increases slightly (less than 1W) when the hardware engine is decoding video instead of the ARM cores. You find a few before/after measurements in the benchmarks thread in 'Linux Development' forum. As usual even cheap heatsinks help already a lot (10-15°C less when the system is busy) and by choosing an intelligent enclosure with a lot of vents convection can also help. But Pine64 won't need a heatsink in most situations since we already fine-tuned throttling settings (we rely on the internal thermal sensor and keep the maximum temperature at 90°C with a bit safety headroom to be able to initiate an emergency shutdown which works pretty well in the meantime but has still some room for improvement) so without a heatsink peak performance will be affected but this is something most people won't experience unless they run benchmarks or try number crunching. (03-29-2016, 07:01 AM)Artyom Wrote: As it turned out, he averaged more heats SOC and RAM than linux without much stress. Then simply play FullHD video. And I realized that it is necessary cooling. Since according to measurements from the Android software that comes SOC RAM and Then that's a driver problem in Android or you chose a format that's not supported by HW decoding (HEVC in 10 bit for example). How did you measure DRAM temperature?
03-29-2016, 07:55 AM
Backer 31,233 (2 x 2GB boards)... March nearly over - still not here...
03-29-2016, 08:34 AM
(03-29-2016, 07:08 AM)Andrew2 Wrote:(03-27-2016, 11:11 AM)Artyom Wrote: 62 C degrees temperature on SOC when viewing full HD movie Just! I have no special means of temperature measurement, has not yet has got. So I just touched the ground first (removed static electricity) then DRAM. By the way. After installing the radiator in my subjective feelings Android began to work a little faster.
03-29-2016, 09:53 AM
(03-29-2016, 08:34 AM)Artyom Wrote: I have no special means of temperature measurement, has not yet has got. So I just touched the ground first (removed static electricity) then DRAM. By the way. After installing the radiator in my subjective feelings Android began to work a little faster. Well, that's also how our mind works (I've spent hours on improving something and now I can even feel that it was worth the efforts ,) ) But normally it should be faster even if it looks you were a bit frugal using thermal paste (given the SoC fits exactly in the white square). If you would've used longsleep's Ubuntu Xenial image you could've used my RPi-Monitor installation script to measure before/after easily BTW: If Android uses the default throttling/killing strategies we found in Allwinner's BSP then you might end up with killed CPU cores instead of safe throttling. But I've no idea how to check this stuff in Android. The same applies to monitoring temperatures and CPU clockspeeds... |