LotofBench : Multi-arch Compute benchmark
Go back to list
LotOfBench v1.0 is the first benchmark of LotofBench, the open source, dependency free CPU suite modelled on the structure of Geekbench 6: identifiable workloads, single core and multi core passes, per workload scores against fixed baselines and a composite built from geometric section means. LotofBench scores carry their own calibration and are not comparable with Geekbench scores; the project therefore runs Geekbench 6 on the same machines to give each LotofBench result its Geekbench 6 equivalent.
The project dissects the results LotofBench reports:
- the Single Score and the Multi Score, the two composites the suite publishes;
- each of the 16 workloads with its explanation, its result and its scores, split between an Integer section (File Compression, Navigation, HTML5 Browser, PDF Render, Photo Library, Clang, Text Processing, Asset Compression) and a Floating Point section (Object Detection, Background Blur, Object Remover, Horizon Detection, Photo Filter, HDR, Ray Tracer, Structure from Motion);
- the Geekbench 6 equivalents, measured workload for workload on the same hardware.
The hardware panel uses three AWS generations in Paris (eu-west 3): the Graviton m9g.2xlarge (arm64), the AMD m8a.2xlarge and the Intel m8i.2xlarge (x86), all eight vCPU and 32 GB.
Every machine runs both suites: LotofBench (single and multi passes at the CPU count and twice the CPU count) and Geekbench 6. The charts follow the Geekbench conventions: the single over multi composite multibar and the 16 axis spider web graphs per workload score.