RPi 4b vs RPi 5 benchmark in BOINC - Asteroids@Home
TL;DR: For Asteroids@Home the Pi 5 is roughly three times faster than Pi 4b.
This post is a continuation of the previous blog in which I compared Pis crunching Einstein@Home, so here I omit the parts that would be identical. For more information on the test background and methodology, please read the previous blog post.
The BOINC project I chose for crunching in this test is Asteroids@Home. This is the second of the two projects that provide tasks for ARM 64 bit boards.
Asteroids@Home task uniformity
In the blog post about Einstein@Home I mentioned that BOINC does not always send task of uniform size, some of them are smaller than the others and may skew test results. Asteroids@Home also has this issue. From what I observed most of the tasks are of a similar size, but there are also some outliers which run time is much, much shorter, around 1/2 - 1/3 of the typical tasks length. Because of that I am omitting the “minimum task time” part in this test, because it would only show whether one of the Pis was lucky to receive such a task.
What is also worth mentioning, is that Pi 4 received tasks called “Period Search Application v102.14aarch64-unknown-linux-gnu”, while the Pi 5 tasks were called “Period Search Application v102.14”. The application version is the same, so I assume there was no difference, but I will validate that claim in the A@H forums and update this part once I have an answer.
An interesting difference between Asteroids and Einstein on the Pi 5 was the thermal load of the CPU.
As in the previous test, my rig has a 120mm fan running at 6 volts, cooling both RPi 4 and RPi 5 in an open cluster case. The Pi 5 also has its dedicated, PWM-controlled active cooler attached.
When crunching Einstein on the RPi 5, the temps were around 55C. The dedicated cooler did not start, the large fan cooled it enough to keep the temps below the threshold that would trigger the active cooler.
Asteroids put a much higher thermal load on the Pi 5’s CPU, the Pi was again at around 55C, but to achieve that temperature, the active cooler was running at 5k RPM the whole duration on the test. Whereas the Pi 4, with only a heatsink attached, did not pass 60C as with Einstein. Seems that Asteroids are pushing the 5’s cores much harder. Maybe that is what is causing the massive change in performance?
Here I have to praise the Pi 5 active cooler. Even though running at 5k rpm, it was inaudible, drowning in the noise of the 120mm fan running at half power.
Let’s see the results.
Average tasks time
For the 4b, the average tasks time was 38010 seconds (10 hours, 33 minutes and 30 seconds). For the Pi 5, the average task time was 12207 seconds (3 hours, 23 minutes and 27 seconds), around three times faster.
Maximum task time
For the 4b, the longest task took 42208 seconds (11 hours, 43 minutes and 28 seconds). For the Pi 5, the longest task took 15244 seconds (4 hours, 14 minutes and 4 seconds), again more or less three times faster.
Average tasks per day
A day has 86400 seconds. Both Pi 4 and Pi 5 have four cores, so they have 345600 core-seconds per day. Dividing that number by the average task time, the results are 9 tasks per day for the Pi4 and 28 tasks per day for the Pi 5, the Pi 5 can crunch three times more tasks per day than the Pi 4.
Benchmark raw results
I am attaching below the benchmark data. Please feel free to use it in whatever way you want, just please link to this blog post if you want to publish your work.
Benchmark data in LibreOffice Calc format
RPi 4b results in CSV
RPi 5 results in CSV
Summary and further steps
This has been the second post in the series about comparing Pi 5 with its predecessor. As I said in the previous blog posts, there are also other scenarios in which I want to test the newest Pi, and now I think the immediate next one will be running Universe@Home.
Thanks for reading!
If you have any comments about the test, its methodology, my calculations, or have suggestions what to test next, drop me an email, or contact me on Mastodon. Links are in the footer.
You can help with funding my future projects by supporting me on these crowdfunding sites: