Operating temperature

Does it really matter when full load of CPU causes minimal difference?
For me, yes. My router goes 99% of time on idle load.

we going to get significantly longer lifetime if we optimize temperature on CPU
Opinions differs. I personally think, yes (you can see that on post above). But its just my “common sense” guess. I think lower temperatures -> longer lifetime in general. 90°C is quite high for me :slight_smile:

60°C idle vs 90°C idle

1 Like

hi all guys

Today I check one of my Omnias and I got 75 - 82°C in idle is it danger for router?

exander77 Oct '16
High? 77°C is nothing. This in not desktop, these ARM CPUs for routers can usually go up to 120-130°C. Better look up the specs.

:wink: It´s ok

1 Like

Hi there,

since I got my Turris Omnia this week I’m interested if anyone did find a solution to get the Sytem Temperature via SNMP to show in a LibreNMS Monitoring System.

Unfortunately the script solution provided by idpitt above doesn’t get the sensor to show up in LibreNMS (although it does show if I query the UCD-SNMP-MIB::extTable).

Thanks in advance

Does anyone, who have replaced the original cooling block, still have it lying around?
If so, can you measure it and post the dimensions?

My Omnia is not physically accessible at the moment, and I don’t know how the get cooler block dimensions via ssh :slight_smile:

I suppose it has the same dimensions as in T1 - 20x30x30 mm

1 Like

My Omnie temperature without internal HW adjustment in a no-load state is 68 degrees Celsius. The room is 22 degrees.

root@Omnia:~# thermometer
CPU:    68
root@Omnia:~#

After 5 minut stress is thermomether 78 degres. The only modification is the location of the case.

Right now my Omnia shows 69°C on idle, without any modification at all, outside temperature most likely very close to yours.

Hi,
I would like to just share my CPU heatsink modification. The target was to stay with passive cooling as I don’t like fan noise and dust.
Temperature after modification is:

IDLE temperature is 56°C

after 20 minutes of CPU stress I reached max. 72°C




One more update :grinning:

2 Likes

Although running at 70C-80C seems to be perfectly legit according to the Armada 38x specs as the chip’s operating temperature is up to 105C, I’ve tried to apply the solution recommended by @backon earlier in this thread just to see the effect.

I used the ARCTIC Thermal pad (1.5mm), few notes below:

  • Removing the cube was really easy as the original thermal tape didn’t hold it too strong, it was just sufficient to push the cube gently sideways. The tape remained completely attached to the cube, which made it’s removal with a screwdriver and alcohol really simple.

  • Then, I’ve cut two 20x20x1.5mm squares from the Arctic pad and applied one to the bottom and the other to the top of the aluminium cube.

  • Finally, I’ve attached the cover back and made sure it sticks to the pad by pressing on the cover in the place where the cooling cube touches the cover.

  • The whole thing was done in a few minutes.

The results are surprisingly good, my Omnia’s temp dropped from the usual 74C to 66C.

This is 8C (~10%) improvement just by applying the pad worth a few dollars. Neat!

3 Likes

I applied the modification described by TomasKapin and Bacon. At home there was about 26°C at the time of change. Before the change Omnia was operating in about 78°C when idle. After setting Arctic Thermal pad (1.5mm) on the top and bottom of aluminium it was about 69°C when idle and increased up to 80°C after 40minutes of stress test.

I picked up the newer revision of the OPOLAR laptop fan (with the temp display) recommended by @hdBerretz (thanks!) for my Omnia and it works like a charm. With the fan at 2/3 speed, it lowers the idle temp by around 11C - air and CPU (with pakon running).

The ambient temp in my flat is around 26C. At idle, the air temp reported by the fan is around 28C; the CPU temp is around 58C.

Regarding acceptable operational temperatures, I realize the CPU can handle rather high temperatures without issue, but I’m concerned for the mSATA SSD, which sits right next to the CPU, and is more sensitive to heat.

I have no idea what the temps are like inside the case under load, but I feel like a CPU running at 100C for any length of time could make the SSD dangerously hot, particularly if the SSD is also under load.

It’s a shame they didn’t put more effort in the cooling solution.
They just put a brick of aluminum on top of the proc. which is only held in place by the cover and some tim. I wonder if this solutions holds when the TO is operated in a vertical way.
If only there were mounting holes to mount a more efficient/elegant cooling solution.

On a side note: it’s gets the job done. Other comparable devices have equally bad cooling solutions or a complete lack thereof.

It does, no issues there.


CPU heat sinks (passive) are likely the common place for SoC nodes, the only other way would be active cooling with fans and that would probably not fit the case.


It should be possible to query that info with smartctl -x

e.g. on this node `smartctl -x /dev/sda | grep Temp` prints

190 Airflow_Temperature_Cel -O–CK 056 021 000 - 44
Current Temperature: 44 Celsius
Power Cycle Min/Max Temperature: 37/45 Celsius
Lifetime Min/Max Temperature: 29/79 Celsius
Specified Max Operating Temperature: 55 Celsius
Under/Over Temperature Limit Count: 0/0
SCT Temperature History Version: 2
Temperature Sampling Period: 1 minute
Temperature Logging Interval: 10 minutes
Min/Max recommended Temperature: 0/70 Celsius
Min/Max Temperature Limit: 0/70 Celsius
Temperature History Size (Index): 128 (124)
Index Estimated Time Temperature Celsius
0x05 ===== = = === == Temperature Statistics (rev 1) ==
0x05 0x008 1 44 — Current Temperature
0x05 0x020 1 79 — Highest Temperature
0x05 0x028 1 29 — Lowest Temperature
0x05 0x058 1 55 — Specified Maximum Operating Temperature

It would seem that the Specified Maximum Operating Temperature had been exceeded indeed but it does not reveal when and over which period of time…

Unfortunately apps like hddtemp or nvme-cli are not available on OpenWrt | TOS.

2 Likes

I have just applied this. I have quite recently hit 118 degrees celsius in big load. During some networking error.

Now I have 60 to 70 idle and 85 during max stress. That is great improvement.


Hmm… last week the temperature went nearly to 120 C. Looking at your cooling tweaks… should I be concerned?

Anyway I decided to take the step anyway. Being unable to remove the block from the board I only applied the pads to its top. Here’s from a preparation:


I had the case opened several times, completely forgot there was a thermal compound in the past :astonished:

And now after fixing it:

Pretty much right after this the temperature went down by approx. 10 degrees C. And also in the long run:

So yeah… I very much recommend this to anyone experiencing temperatures of 90 C and above.

Room temperature 23 °C. According to the instructions @bacon installed today --> 72 > 65 °C
DAY
image

HOUR
image

HOUR first stress — 85-87 °C
image