I’ve been trying out a few things with a new Raspberry Pi 5.
I have a small test app that I use with pro bono “customers”. The results are for a simple web front-end loop test app running locally to load data into SQLite, based loosely on Dana Brown’s web Task SQLite Example. The idea was for the test to cover common SQL operations.
WAL is turned on and I have tested with up to 9 separate client front-ends on a WiFi LAN.
Each loop inserts 2 rows into a table, and 3 triggers insert a row into a log table for each added/updated/deleted main table row. The loop updates each added main row; deletes a row from main table and retrieves the last 3 added rows from the log table.
The final results for 12,000 rows added/updated/deleted in the remote debugger were:-
Raspberry Pi 5 microSD Bookworm: 6,000 : ran at 53.8 rows/sec from two clients
Raspberry Pi 4 microSD Bookworm: 6,000 : ran at 39.9 rows/sec from two clients
Implying that the Pi 5 was about 35% faster than the Pi 4 - This is not the whole story as disk speed is a significant bottleneck (A faster disk gave ~260 rows/sec).
General thoughts on Raspberry Pi 5:-
Using a WD 2TB HDD boot disk, the system would not mount with the standard Pi 15W power supply. It did mount with 40W, but required a high-spec USB cable - The Pi 5 requires the 25W (or better) power supply if used with an external HDD.
Uses more power than the 4B, and runs (much) hotter. Another app caused the Pi to hit 86C and crash.
An Active Cooler or efficient case is essential for high loads.
The built in on/off switch is useful.
Requires Bookworm, which I found to be a “work in progress”.
For now, I will stick with Bullseye on the 4B. I might buy a Cooler (fan noise?), but FLIRC will ship a 5B case soon. I hope to re-evaluate Bookworm and the Pi 5 again later. YMMV…
Caveat: Bookworm uses more systemd stuff, I’m not a fan - I was using BSD style *NIXs in the 1980s - It seems to me to be against the standard *NIX philosophy, so I may well be biased…