Raspberry Pi 4 Benchmarks – Heatsink Edition

A few days ago, I ran some benchmarks in Raspberry Pi 4, and quickly found out that using the board without a cooling solution will cause serious performance issues, as in some cases my board was slower than Raspberry Pi 3 model B due to severe overheating.

After playing with LibreELEC yesterday, I’ve now reinstalled Raspbian Buster Desktop on the board, and fitted it with a largish heatsink and some old thermal paste.

Raspberry Pi 4 Heatsink
Click to Enlarge

So I’ll run benchmarks again with and without heatsink. I’ll only run sbc-bench this time.

SBC Bench Installation

Open a terminal windows or connect to the board through SSH and run:


Raspbian Buster will automatically fetch the latest operating systems packages upon first boot, but apparently not the latest firmware:


So I ran rpi-update to get the very latest firmware as well, and rebooted the board:


Normally, you should not have to do it, but Raspberry Pi 4 is really new, so I expect frequent changes at the beginning. Read warning in rpi-update:

WARNING: ‘rpi-update’ updates to pre-releases of the linux
kernel tree and Videocore firmware.

‘rpi-update’ should only be used if there is a specific
reason to do so – for example, a request by a Raspberry Pi
engineer.

Results with heatsink

Room temperature: 28°C. Uptime, idle temperature, and CPU freq:


Benchmark results:


Frequency/temperature monitoring during 7-zip:


It never went over the maximum 85°C temperature, and always under 80°C.

Results without heatsink

Now let’s remove the heatsink, reboot the board and wait a few minutes.

Room temperature: 28°C. Idle temperature:


So 4°C more at idle after about 3 minutes uptime without heatsink..

Benchmark results:


Throttling did occur, and indeed monitoring frequency and temperature during the benchmark shows temperature going close to 85°C, and real frequency dropping as low as 750 MHz to cool the system:

Comparison table and Pretty Chart

We can see that single thread benchmark are not affected by the presence of an heatsink, but the multi-threaded 7-zip compression is certainly impacted.

BenchmarkRaspberry Pi 4 “Naked”Raspberry Pi 4 “Heatsink”Ratio
memcpy2608.82540.197.37%
memset3715.63541.895.32%
7-zip44235397122.02%
OpenSSL
aes-256-cbc 16KB
64891.56k64782.34k99.83%

The differences for memcpy, memset, and OpenSSL are just statistical errors. But Raspberry Pi 4 with heasink is over 20% faster for 7-zip compression as it does not throttle during the test.

Raspberry Pi 4 Benchmarks Heatsink
Click to Enlarge
Support CNX Software - Donate via PayPal or cryptocurrencies, become a Patron on Patreon, or buy review samples
Advertisements
Subscribe
Notify of
guest
25 Comments
oldest
newest most voted
Advertisements