0

I have a Raspberry Pi 4B 8GB running non-stop under a fairly mild load. Among other things, it serves a somewhat large PostgreSQL DB, but the mean CPU usage rarely exceeds 25%. There is a Low-Profile ICE Tower cooler mounted onto the SoC. The board is running the latest version of Raspberry Pi OS (32-bit, desktop at 1920x1080, no overclock).

While the cooler functions perfectly well, I am a little bit worried about how long it can keep on running non-stop. I know from experience with various laptops and desktops that even the larger fans wear out after a couple of years of constant spinning and I assume it is even worse for the small fans. Therefore, I am trying to find the easiest solution that would prevent unnecessary wear of the cooler's fan without compromising the SoC's life expectancy. The most obvious way of achieving the former is by simply unplugging it. However, this leads me to wonder if I am, perhaps, worsening the latter.

I understand that the fans are made to spin, but in the case of SBCs I prefer the aesthetic and simplicity of passive cooling, where possible.

For reference, here are my SoC temperature readings. Please, do take them with a grain of salt. I did not really go through any rigorously scientific process of recording the values, so I am eyeballing the values here a bit.

Room Temperature: ~20°C

  • With 5V Fan: 26-29°C
  • With 3V Fan: 29-32°C
  • Passive Cooling Only: 40-45°C

Room Temperature: ~30°C

  • With 5V Fan: 30-33°C
  • With 3V Fan: 34-37°C
  • Passive Cooling Only: 45-50°C

At the time of writing this question, the room temperature is 33°C and the SoC reports 46°C without active cooling. This is after running for slightly over 3 weeks without a reboot, last 6 hours of which have been without a fan.

It is quite possible that the temperature could temporarily jump up to 55-60°C during a spike, but I do not expect that to happen very often, nor should it happen for prolonged periods of time.

This, finally, leads me to my question:

Is it safe to run a Raspberry Pi 4B non-stop for a very long time (several years) without a reboot at 45-50°C? Would it be significantly better for the SoC to run at 35-40°C instead?

I am aware of a similar question, but I believe mine to be somewhat different.

natiiix
  • 131
  • 5
  • Consider [this answer](https://raspberrypi.stackexchange.com/a/114457/83790), and [this answer](https://raspberrypi.stackexchange.com/a/84212/83790). These may help with perspective. – Seamus Aug 21 '20 at 18:25
  • @Seamus Coincidentally, I read both of the articles linked in the first answer before I bought the RPi to know if an active cooler is necessary for RPi 4 because I always ran the older RPis passively cooled. – natiiix Aug 21 '20 at 20:00
  • Reading is good. – Seamus Aug 21 '20 at 20:11

1 Answers1

2

"Is it safe to run a Raspberry Pi 4B non-stop for a very long time (several years) without a reboot at 45-50°C? Would it be significantly better for the SoC to run at 35-40°C instead?"

Your question invites speculation and opinion. That's not to say it's an irrelevant question - only that it's couched in non-specific terms. I'll try to provide some facts which you may consider in formulating your path forward.

"Safe"

In the context of "thermal safety", the answer is: YES

This due to the fact that the RPi4B design includes a closed-loop thermal management system that regulates component temperatures. This system prioritizes thermal management over performance, and consequently keeps your RPi4B safe.

I am not clear at all on why you feel reboots are relevant here.

"Better"

In the context of reliability or longevity, the answer to your question is: YES

The Arrhenius equation tells us that electronics will "live longer" at lower temperatures. If the math bores you, you can easily find articles similar to this one that explain this relationship, and propose various rules of thumb to estimate the effect of temperature on component life.

I will leave it to you to determine if that is significant to your use case.

Seamus
  • 18,728
  • 2
  • 27
  • 57
  • Thank you for your insight! I was aware of the questionable word choice with "safe". I meant "not unhealthy for the SoC in a significant way" / "not worsening its performance". I would consider thermal throttling as an unsafe state in this context. I mentioned reboots to emphasize that there are no breaks in the workload - it starts once and never stops until either the project or the platform dies. A reboot (or rather a shutdown) could indicate a period (no matter how short) of being completely idle. Seems like 50°C would *probably* be okay, but even a fan like this is cheaper than a new Pi. – natiiix Aug 21 '20 at 19:55
  • @natiiix: Throttling is not "unsafe" behavior... I'm completely puzzled why you would say that. Perhaps you've left something out of your question? – Seamus Aug 22 '20 at 21:31
  • It's not "unsafe" in the sense of having the potential to harm the hardware. However, it's a clear indication of insufficient cooling and generally happens at the edge of what's considered a safe operating temperature (i.e., in desktop/laptop graphics cards). Even if it didn't indicate any such issue, it's undesirable to compromise the performance, which is precisely what thermal throttling would result in. Therefore, while protecting the hardware, it hurts the workload, which would be a much bigger problem for me than a worn-out fan. – natiiix Aug 22 '20 at 21:49
  • @natiiix: I think you are confused... I'll leave it at that. – Seamus Aug 22 '20 at 21:53
  • I started my first comment by saying that crippling the performance constitutes an unsafe, or rather an undesirable, state, which is consistent with thermal throttling being hugely undesirable. You certainly wouldn't want to run a critical server in a thermal throttling state for years and call that safe. – natiiix Aug 22 '20 at 22:46
  • @natiiix: Yes, actually I would, and without hesitation. The only exception would be a task with two attributes: 1) Demanded near 100% CPU/GPU capacity for extended periods of time, AND 2) Had real time or *near* real time requirements. Your job may fall into one of these categories. If that's the case, consider more computing "horsepower". However, as your question includes no measurements showing the "supply vs demand" metrics *suggests* that you don't understand how this works. If I'm wrong, pls excuse me for being forward. In any case - I'd like to drop this conversation. – Seamus Aug 22 '20 at 23:09
  • Pushing hardware into a thermal throttling state simply seems like redlining a car to me. Sure, it's not an issue if it happens during a spike, but running the hardware like that for years? It's obviously a state it doesn't want to be in, considering it's a safety precaution in itself. As you indicated yourself, higher temperatures lead to a shorter semiconductor life and the temperature at which thermal throttling begins to happen is at the edge of what the hardware itself considers safe, so it doesn't feel like something that'd be okay for a long time. Regardless, thanks and bye. – natiiix Aug 23 '20 at 17:20