benchmark Broadcom BCMxxxx Debian Raspberry Pi

Raspberry Pi 4 Benchmarks – Heatsink Edition

Orange Pi Development Boards

A number of days ago, I ran some benchmarks in Raspberry Pi 4, and shortly came upon that using the board and not using a cooling answer will trigger critical efficiency issues, as in some instances my board was slower than Raspberry Pi 3 model B as a result of extreme overheating.

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

Raspberry Pi 4 HeatsinkClick to Enlarge

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

SBC Bench Set up

Open a terminal home windows or hook up with the board by way of SSH and run:

Raspbian Buster will routinely fetch the newest working techniques packages upon first boot, however apparently not the newest firmware:

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

Normally, you should not need to do it, however Raspberry Pi 4 is admittedly new, so I anticipate frequent modifications originally. Learn warning in rpi-update:

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

‘rpi-update’ ought to only be used if there is a specific
purpose to do so – for instance, 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 throughout 7-zip:

It by no means went over the utmost 85°C temperature, and all the time beneath 80°C.

Outcomes with out heatsink

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

Room temperature: 28°C. Idle temperature:

So 4°C extra at idle after about 3 minutes uptime with out heatsink..

Benchmark outcomes:

Throlling did occur, and certainly monitoring frequency and temperature in the course of the benchmark exhibits temperature going near 85°C, and real frequency dropping as little as 750 Mhz to cool the system:

Comparison desk and Fairly Chart

We will see that single thread benchmark are usually not affected by the presence of an heatsink, however the multi-threaded 7-zip compression is definitely impacted.

BenchmarkRaspberry Pi 4 “Naked”Raspberry Pi 4 “Heatsink”Ratiomemcpy2608.82540.197.37%memset3715.63541.eight95.32%7-zip44235397122.02%OpenSSL
aes-256-cbc 16KB64891.56okay64782.34okay99.83%

The differences for memcpy, memset, and OpenSSL are just statistical errors. But Raspberry Pi 4 with heasink is over 20% quicker for 7-zip compression because it doesn’t throttle in the course of the check.

Raspberry Pi 4 Benchmarks HeatsinkClick to Enlarge