Does the Pi4 really run hotter then the Pi3?

When the Pi4 was release last year, many people complained about the higher power consumption and higher CPU temperatures compared to the Pi3. We did some measurements and could see that power consumption and therefore heat dissipation was much higher than on the Pi3B+.
But lots of development have been done in the last 12 months and power management on the Pi4 have been greatly improved. Let’s just have a look how the situation is today.

For this test, we’re using the following test setup: The Raspberry Pi will be equipped with a relatively large heat sink. We are using the latest version of HiFiBerryOS based on Linux 5.4 (this is not yet released, we’re still in the process to test everything). It plays a web radio stream. Network is connected via 100 MBit Ethernet.
CPU temperature is measured using

cat /sys/class/thermal/thermal_zone0/temp

Power consumption is measured on a bench power supply that provide stable 5.1V to the Pi.

Why are we using the heat sink? First of all, many users do this. But there is also a technical reason: The Pi’s SOC uses a heat spreader that is highly reflective. That makes thermal imaging quite hard is reflective surfaces will show lower temperatures in thermal images (you see in the thermal images below that connectors seem to be much cooler than other parts, but as least part of this is caused by the reflective surface).

Pi3 B+

CPU temperature: 47°C
Power consumption: 2.1W


CPU temperature: 48°C
Power consumption: 2.2W

Why is the difference in the thermal images 2°C instead of just one in the measurements of the CPU? The heat sink that we’re using here wasn’t designed for the Pi3. There probably isn’t the best thermal connection between the Pi’s CPU and the heat sink. That means some of the energy will be dissipated to other parts, e.g. the PCB. This is hard to see in the thermal images.

A Pi4/Amp2 combination

But what about a “real life” system? The examples above don’t have a HAT on top of the Pi and they are not mounted in an enclosure. Let’s have a look at a real system. We’re using our office system: A Pi4/Amp2 combination. It’s also running HiFiBerryOS and playing a radio stream. There is a pair of Tannoy System 8 speakers connected. Volume is measured at 72db peak at 1m. This is already much louder than our normal background listening volume. We’re using a 20V power supply here.

CPU temperature: 63°C
Power consumption: 4.6W

The CPU temperature is a lot higher here. The difference to our previous measurements is caused by the following:

  • There is a board on top of the Pi that makes heat dissipation harder
  • Heat dissipation is only possible via the steel case
  • the Amp2 creates additional heat

Just note that 63 °C is not a problem for the CPU at all. It is designed to work even at much higher temperatures. The case is just warm at 38 °C.


In low-load use cases (as most audio use cases are), there is almost no difference between a Pi3 and a Pi4 anymore. Power management has been greatly improved within the last 12 months.
While we still recommended the Pi3 for some use cases in the past, we would not do this anymore. The additional power consumption of 0.1W and the slightly higher temperatures don’t make
a difference in real-world applications.

Note that in use cases with high CPU load, the Pi4 will still run hotter and consume more energy than the Pi3, but it is also more powerful and can handle loads that the Pi3 can’t.

September 23, 2020

Subscribe to new blog posts