All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
nerdbench - open source, single-binary CPU benchmark
Introducing nerdbench, a simple open-source alternative to Geekbench
It does the boring job of benchmarking a machine's CPU. it's packaged into a single binary just like geekbench.
It's basically a wrapper around a range of handpicked benchmarks from different categories in the Phoronix test suite. The standard run profile finishes in under 5 minutes, but you can fully customize the runtime. The full source and build pipeline are in the Github repo.
Rationale - after the various Geekbench outages and intermittent upload errors over the last few months, I wanted to put together an open-source alternative that "just works" locally. Plus, I dislike how the free version forces you to upload benchmarks to their site. It's unreliable (as already shown), and is just bad unnecessary telemetry. So, Nerdbench works entirely offline.
Despite the name, it's obviously not a full replacement for Geekbench. But for the average guy who gets a new box and wants to test the CPU, it's likely enough. (And it's faster, since you can customize the run time)
Of course, I realize this is probably not of interest. I'm just putting it up here.
Some things are not yet set up or documented fully. For example, only x86_64 binary is available. I'll finish it soon.
Disclosure, maybe important to you: LLM did the bulk of the work here. Mostly, it handled routine tasks like getting each project to compile and stitching things together. There is almost no logical complexity to the project, as it's all on top of existing libraries and test suites.
drop your thoughts or run results below…?
Run it with
wget https://raw.githubusercontent.com/XInTheDark/nerdbench/main/scripts/bench.sh
bash bench.sh
(there is also one liner, but on LET it is of course blocked)
Please inspect the source before you run! The whole project is transparent so I highly encourage you to audit it
More usage options and source code: https://github.com/XInTheDark/nerdbench
Example output:
nerdbench-linux-amd64: OK
[1/16] sysbench single running... threads=1
[1/16] sysbench single done: 6721.65 eps, score 997
[2/16] sysbench multi running... threads=4
[2/16] sysbench multi done: 24420.62 eps, score 1043
[3/16] c-ray single running... threads=1
[3/16] c-ray single done: 6868507.86 rps, score 1031
[4/16] c-ray multi running... threads=4
[4/16] c-ray multi done: 22295433.23 rps, score 1018
[5/16] stockfish single running... threads=1
[5/16] stockfish single done: 1795467.00 nps, score 999
[6/16] stockfish multi running... threads=4
[6/16] stockfish multi done: 5868581.00 nps, score 948
[7/16] sqlite-speedtest single running... threads=1
[7/16] sqlite-speedtest single done: 6.03 tps, score 5021
[8/16] sqlite-speedtest multi running... threads=4
[8/16] sqlite-speedtest multi done: 19.05 tps, score 5181
[9/16] openssl-speed single running... threads=1
[9/16] openssl-speed single done: 149439.10 ops, score 984
[10/16] openssl-speed multi running... threads=4
[10/16] openssl-speed multi done: 573961.82 ops, score 1065
[11/16] zstd single running... threads=1
[11/16] zstd single done: 12376.96 MB/s, score 1362
[12/16] zstd multi running... threads=4
[12/16] zstd multi done: 14950.17 MB/s, score 1375
[13/16] ggml-ml-kernel single running... threads=1
[13/16] ggml-ml-kernel single done: 31.11 ops, score 995
[14/16] ggml-ml-kernel multi running... threads=4
[14/16] ggml-ml-kernel multi done: 72.16 ops, score 955
[15/16] tinycc-compile single running... threads=1
[15/16] tinycc-compile single done: 2106.56 files/s, score 1266
[16/16] tinycc-compile multi running... threads=4
[16/16] tinycc-compile multi done: 5324.08 files/s, score 1182
NerdBench v2026.06.3 (quick)
Single Core: 1311
Multi Core: 1309
sysbench single 6721.65 eps score 997
sysbench multi 24420.62 eps score 1043
c-ray single 6868507.86 rps score 1031
c-ray multi 22295433.23 rps score 1018
stockfish single 1795467.00 nps score 999
stockfish multi 5868581.00 nps score 948
sqlite-speedtest single 6.03 tps score 5021
sqlite-speedtest multi 19.05 tps score 5181
openssl-speed single 149439.10 ops score 984
openssl-speed multi 573961.82 ops score 1065
zstd single 12376.96 MB/s score 1362
zstd multi 14950.17 MB/s score 1375
ggml-ml-kernel single 31.11 ops score 995
ggml-ml-kernel multi 72.16 ops score 955
tinycc-compile single 2106.56 files/s score 1266
tinycc-compile multi 5324.08 files/s score 1182
Comments
Thank you AI overlords
If you need to calibrate sysbench test results, you can use the values for various CPU types compiled from unofficial sources at https://sysbench.spiritlhl.net/
X in the dark. Nothing can go wrong here.
enjoying last few subsidized tokens before falling into the permanent underclass…
Thanks claude