Skip to main content


Raspberry Pi4 continuous selfhosted server operations


!Selfhosted For Raspberry Pi4 continuous selfhosted server operations: with and without case fan. This graph show a 20°C decrease, with a slow rpm fan.
The metal case has thermal stickers to reach for main components on both sides, and the fan is what I have in scrap parts, totally not for that case but pushes a lot of air with low noise compared to screaming mini cpu fans.


#selfhosted #raspberrypi #homelab

catloaf doesn't like this.

in reply to Ⓜ3️⃣3️⃣ 🌌

No, the highest temperature I have ever recorded was 76°C. Without fan, and using thermal pads to connect the case to components.

But I hate high numbers, excessive heat already shortened a Pi lifetime here, since then I am very cautious about that.

in reply to Ⓜ3️⃣3️⃣ 🌌

I've kept a raspberry pi 4b that's given a mild OC to 1900Mhz in my boiler cupboard for a year and all its needed to keep it below 50 is:

  1. a tiny metal heatsink
  2. a 5w usb fan blowing already-warm air sorta towards it
in reply to Lucy :3

TBF 0C and 0F are kinda arbitrary values.
This entry was edited (18 hours ago)
in reply to WolfLink

0 F is defined as the lowest temperature a guy was able to reach with some random mixture which is arbitrary.

But 0 C is commonly defined as the point at which water turns from solid to liquid and the other way round. Scientifically it's ever so slightly off but still it's defined via Kelvin.

in reply to einkorn

Celsius is defined at the arbitrary standard pressure, different pressure values change how water acts (this is very noticeable at 100C - boiling). Kelvin steps are defined from Celsius. Even the speed of light which is constant is still defined in terms of an arbitrary time unit.
in reply to bluGill

Wow wow guys... that's just RRDTool doing its thing with graphs 😆
in reply to bluGill

Seconds are clearly defined by one specific caesium-133 transition

The second [...] is defined by taking the fixed numerical value of the caesium frequency, ΔνCs, the unperturbed ground-state hyperfine transition frequency of the caesium 133 atom, to be 9192631770 when expressed in the unit Hz, which is equal to s−1
in reply to exu

Caesium might not be arbitary but 9192631770 is.
in reply to bluGill

In the world of embedded programming that would be a

/* Magic number do not change this value */
#define CAESIUM_FREQ 9192631770

in reply to Ⓜ3️⃣3️⃣ 🌌

So how did you attach that fan to the case? Looks like an interesting idea seeing that its powered through the USB.
in reply to ddash

The fan is just sitting on top of the case, mostly centered as the turbine opening is slightly smaller than the case itself.
The duct shoots air horizontally in the general direction of a LibreComputer Lepotate. I take that as a bonus.
in reply to Ⓜ3️⃣3️⃣ 🌌

That is one beefy fan for a rpi.

I've just got a case similar to this; but all snap together, no screws:

The fan runs off the pin headers. Meant for 5v, but I use the 3.3v line to run it a little quieter/slower.

Even that makes a good 10°c difference.

in reply to Ⓜ3️⃣3️⃣ 🌌

I had multiple fan failures which brought down my system, so instead I picked up one of these heatsink cases and it's been running mostly uninterrupted for years.
in reply to Ⓜ3️⃣3️⃣ 🌌

I've got the same Geekworm cases for all my Raspberries. Added some extra thermal pads in some strategic places and have them all without any active cooling at room temperature (20-21℃). They barely go over 55℃:

The 3B+ has soft-throttling (can be disabled) starting at 60℃, full throttling at 70°C. The 4 starts throttling at 80℃ and the 5 at 85℃. So, with that completely passive cooling I'm still far away from these margins. No need for any moving parts that will make noise at some point in the future.

Also, since the cases have ribs on both sides, I have my Pis standing upright to hopefully make use of the stack effect - which might contribute to the slightly lower than OP's values.

in reply to Ⓜ3️⃣3️⃣ 🌌

The CPU is perfectly happy sitting at 50°C. It is slightly happier at 30, but it doesn't actually help in any way unless you run into throttling, or run (much) hotter for longer. It's fine.

Some might state that the CPU is probably gonna live longer, but seriously have you ever had a CPU die on you cause it was old (or even die at all, even)? Again, it's fine.

Having something that mostly agitates the air (not even really moving it) like a low-hundreds-rpm fan would also work. As would using one of those passive heat pipe coolers that are also overkill (especially with a fan, but just leave that off), but have the same "number looks better" effect.

in reply to Creat

When I said I had one Pi lifetime that was shortened it was a way of saying yes, it died from 24/7 high load no fan overheating. Now it just boot and freezes after a couple of minutes.
Then, yes, back in the days when overclocking and over powering was a thing, CPU actually died from all that, I changed a few for customers that didn’t know better. How ironic I did that to myself years later with little arm boards, isn’t it?
in reply to Ⓜ3️⃣3️⃣ 🌌

That's what I said with "much hotter for longer". If it's constantly thermal throttling, that's gonna be an issue. Of course OC'ing also will. 50°C just isn't an issue. Also older models have CPUs that either don't throttle at all, or do it less well/effectively.