Hi, thanks for this, interesting test. I tried it with the HC1 and there was some variation (disk access?), however I found with an SSD it was quicker:fvolk wrote: ↑Tue Aug 13, 2019 2:50 pmFor fun, I use a mixed workload to get an overall first feel for a platform:
Compiling ripgrep from scratch and warm cache (cargo install -f ripgrep; time cargo --offline install -f ripgrep)
That's a single threaded (linking), multi threaded (compile) and I/O (lots of intermediate files) mixed workload using everything a bit.
Numbers for Rust 1.36+ripgrep v11.0.2:
0m43s -- 6-core desktop x86 (nvme SSD)
2m42s -- H2 (nvme SSD)
~6m40s-~7m02s -- HC1 (mechanical disk)
~13m20s -- C2 (microSD)
where there is quite some variance with HC1 (due to thermal throttling?).
HC1, SSD, Armbian, root FS on SSD, in “stack” of 6, cores at 1.8/1.3 GHz (?mostly)
With fan: 4m35s to 5m17s — temps ambient ~20, idle 31, peak 67
Without fan: 4m43s to 4m52s — temps ambient ~20, idle 53, peak 101
I was watching the temp as well out of curiosity...
A few more -
N2, Armbian desktop, 8GB eMMC, 1.9/1.8 GHz
8GB eMMC: 4m37s — temps ambient ~17, idle 30(?), peak 45
32GB SD card: 4m33s — temps ambient ~17, idle 30(?), peak 44
RPi4, Raspbian Buster desktop, no heatsink, no fan, no case, 1.5 GHz
7m03s — temps ambient ~17, idle 56, peak 79
NanoPC T4, FriendlyDesktop, 1.8/1.4 GHz
6m52s — temps ambient ~19, idle 40, peak 70
With fan: 6m41s — temps ambient ~19, idle 33, peak 49
NUC8i5BEH (i5-8259U), Linux Mint, 3.0 - 3.8 GHz
1m02s — temps ambient ~19, idle 45, peak 91