Results for Radxa X4

[geerlingguy]

Using stock install of Ubuntu 24.04 Desktop, and running with the official Radxa X4 Heatsink case, here are my results:

https://0x0.st/XO8j.bin

I believe the CPU hits thermal limits as it's hitting 95°C very quickly on any benchmark run, so I may work on a more robust cooling setup, to make sure that's not the problem with some of the lower performance I'm seeing compared to other N100 systems (and not just whatever power profile is set up).

I saw there was an 'X4L' run posted in this Radxa community forum post but I'm not sure if that's an entirely different hardware product, or was a typo for the X4...

More of my benchmarking/testing here: geerlingguy/sbc-reviews#48

[ThomasKaiser]

There's definitively two issues (and maybe a third called 'power cap' to be revealed after fixing the 1st):

• thermals
• only 32-bit DRAM access (see the 'DIMM configuration:' section of https://0x0.st/XO8j.bin vs. https://pastebin.com/raw/Z2E9WS5e)

This does not only lower memory bandwidth but adds also to memory latency as such 7-zip compression scores are more affected than decompression (directly comparing to this X4L thing that is able to access DRAM at 64-bit width)

[geerlingguy]

I've re-run Geekbench and HPL, and both are marginally improved (like 2-3% gain), but I've also re-run sbc-bench.sh and here is the result: https://0x0.st/XObB.bin

[geerlingguy]

Not much difference (slight improvements across the board), but temps were down about 10° using paste instead of pads, even with a notable gap, about 1mm or maybe 0.6mm between the heatsink metal and the die.

[ThomasKaiser]

Closed by 76420ff