Temperature thread

Recently I get touch with new boxes, which has nighlty NG kernel friendly SOC. I did some tests to find out, how to reduce temperature on chinese cheaper ones. Here are my results with current nightlies, it could help someone to evaluate his own temp, if its normal, or not, and if he could do something to lower it down.

Hopefully, PR will get merged at some point.

UPDATE: Ondemand governor is now default governor also on ng kernel builds.

Feel free to use this thread also for other SOC’s, if you have done some testing.

I am planning to tune X96 Air P3 a bit more, so if I will have some new findings, I will update this accordingly.

More conditions could affect temperature, as used addons, NAS shares etc. (for example, Odroid C2 have 43 idle on my home living room, but 37 on two other installations also on living rooms, which I did to my relatives). So dont take this values as the only correct, but relativelly, the differences between devices in the list should be correct, as I try to keep conditions same.

P.S.: Room temperature 22 °C

One very important bit of information when doing any temperature comparisons is knowing what the room temperature is at the time you are reading your other temperature measurements. Not what the thermostat is set to either but an actual measured temperature taken within a couple feet of the device.

1 Like

In my case it would be 22°. All devices except Rpi2 was tested in the same room/spot.

Very good point.

When my Airconditioner is on , My hardware can be 5 to 10 celcius cooler when doing a read.
The air movement around the TV and box can be felt when going up close.
Even though the room temp may be 23 celsius in the room whilst running the Airconditioner the same 23 celsius on a different day without the Airconditioner will give higher temp readings .

Chaps, I have mixed feelings about posting this. Does the temperature really make that much difference?

I ask because my N2 with super heatsink (apparently) is at 44c whereas my H96 pro S905x2 is at 41 both in rooms about 21c when watching BbC in HD over TvHeadend.

That said, I’ve a friend who lives in Sri Lanka who seems to be able to blow boxes up in no time. I’ve never lost a box due to heat - bricked a couple cause of attempted firmware changes, but he’s been thro 3 Nexteon red tick S805’s that I thought were bullet proof. Could their base line ambient temperature of sweat rolling hot and humid be the cause?

Sorry for the ramble but opinions, as valued posters, you are welcome.

Unfortunately the onboard temp sensors are not particularly accurate.

Humidity is likely to make reliability much worse as there is a chance of electrical shorts from, condensation. In the old days they used to produce “tropical grade” electronics to account for this.

However the main failure mode for electronics is on and off cycle heat stress where everything expands and contracts and stresses the solder joints. The higher the operating temp the worse this is and with heat flow surface mount components this could be significant. The reality for most people is highest reliability is achieved by leaving things running all the time.

Shoog

1 Like

The temperature doesn’t really make any difference.

As for your comparison between the N2 and H96 pro. You can’t really directly compare temperature between the two since they have significantly different SOC. The N2’s s922x has 6 cores, 4 a73 and 2 a53, this will generate more heat then the H96’s s905x2 with it’s 4 a53 cores soc.

It’s kind of like comparing gas mileage of a little 2 door car with a small engine, with a larger sports sedan or a truck with a v8 engine.

My N2 runs significantly cooler than my X96Air even after I added cooling to the later.

My X96Air runs perilously close to throttling temps and it is never asked to do anything particularly demanding. I have my doubts it would cope with sustained software decoding of 1080i files.

Not to say that these boxes cannot happily run stock without hitting thermal issues, but its just out of my own personal comfort zone and I felt I wanted to add extra cooling.

Shoog

I wouldnt be worried. As cdu13a wrotes, they have different number of CPUs.

I assume, you are running performance default governor on N2, so all of those CPUs are always on max speed. That would make this difference easilly, even when both boxes do “nothing”.

I wouldnt be worried about temps around 44 anyway - its fine. If you want to “play”, you could try other governors. (although I dont see that kind of impact on 9.2.0 with module ondemand, on the new NG with ondemand in kernel I see significant difference between performance and others)

Hi,
Looking at netdata the CPUs seem to be in WFI 90% of the time I don’t know how much power it saves on the stm32f103 it makes a lot of difference not as much as a full sleep but normally the CPU stops until the IRQ.