I was disappointed to hear that the V3 mangoES hardware has a fan that apparently runs all the time. If it weren't for the fan there would be no moving parts. The move to a model that required a fan must drastically reduce the mean time before failure.
v2 was fanless v3 has a fan?
I hear you. We weren't thrilled that the model of the main board in the ES moved to having a fan, but there are some options!
The fan speeds and temperature threshholds are set in a udev rule on the ES, in /etc/udev/rules.d/60-odroid_fan.rules . The set of four numbers starting with 33 are the percent of the pwm to run the fan at, and the three numbers starting with 55 are the temperatures to transition between the fan speeds. If you're in a cold environment, you can turn the fan off unless the temperature goes high (a value of 0 as the first of the four numbers will turn the fan off in the lowest temperature range). Both sets of numbers must be increasing. There is a hard limit around 95-97 C where the board begins clocking itself down, and can select to shut off due to temperature around 100C.
If you want to just experiment with fan settings, you can
cd /sys/devices/odroid_fan.13and then
echo "36 37 38 39" > fan_speedsas this is basically what the udev rule is doing, but without having to restart or make udev re-apply the rules.
The settings that are there to begin with are the product of me tuning down the fan settings from the manufacturer, then back up after experimenting with running a MangoES in my freezer. The original fan settings were obnoxiously loud. But, if you turn down the bottom setting to less than 33 the fan can have real trouble maintaining spinning in a frozen environment. I judged that to be the lowest sensible default. A default of 0 would lead to a lot of switching on and off for the average user, where a default of 20 will make a horrendous beeping in freezing temperatures.
It is possible that this default isn't the true ideal fan setting for any one user. People deploying to cold climates could consider dropping the first speed to 0 and bumping the second speed to 40 or 45. People deploying in warm or hot climates may be satisfied lowering the bottom two speeds for quiet, power, and perhaps longevity, perhaps to 20 and 30
Out of curiosity, how do you expect the fan to fail? A component failure or too much dust will cake into the motor and it won't be able to turn anymore?
It should also be noted the manufacturer of the board does now (not when the V3 was being designed, though) offer a heatsink option, but it is quite large. The reason the V3 requires a fan while V2 does not is that the board is much more powerful.
Thanks for the details. What is the highest CPU temperature that is OK before the fan starts? The ambient temperature in the control panels is 20-35 degrees celcius because the control rooms are climate controlled.
The V2 has a cpu temp of 53 degrees with a consistent load: load average: 6.06, 2.96, 1.29. What temperature does the v3 need to turn the fan on? I don't have easy access to the projects at which we deployed the V3s since they have been turned over to the customers so I can't easily check.
If the heat-sink option were available we would prefer it. Size and noise aren't really a concern for us since we are mounting the mangoES in a control panel 1m deep, 2.5m tall, 3-4m wide we have lots of space and the plant is loud anyway. The area outside the control panels is often extremely dirty as the control panels are installed while the plant is just finishing construction, so everything gets a coating of concrete dust kicked up from sweeping the floor. If there is anything moving the air around it turns in to a dirt pump.
Dust and bearing failures is what I have seen kill fans before. Some fans are really high quality and will probably last long enough but it is difficult to know how good the fan you are getting is going to be whereas if there is no fan at all then it is once less thing to worry about. There must be a good reason the defining characteristic for industrial PCs is often that there are no moving parts.
we have at least 7 projects coming up next year that will get a mangoES. let me know if there is any way we can get the fanless option.
I'll check back in to this thread in 5-10 years and let you know how the single fanless V2 we have deployed vs. the V3 with fan (and hopefully V3 without fan) have held up.
I wrote the default such that the fan is always on, and it's on at the lowest power level so as to behave okay in temperatures down to about -10C (a freezer). You can configure this in the udev rule as I was explaining. You could configure it such that the fan basically never turns on unless thermal failure is imminent (the firmware clocking down or shutting the device off). To do so would be to set the fan speeds to "0 90 93 95" and the temperatures to "80 85 90".
This means PWM 0 until 80 C, then PWM 90% until 85C, 93% until 90C, and 95% thereafter. The firmware starts getting upset around 95C, and gets catastrophically angry around 102C.
I'll certainly ensure your concerns are heard, and let you know if we produce a fanless option, or council you on how you could go fanless.
The bit you mention about industrial PCs not having moving parts is why we weren't thrilled about it either.
any fanless hardware on the horizon?