Lab tool · macOS
MacBenchmark
Native benchmark to certify Apple Silicon MacBooks: measures CPU, RAM, SSD, network, battery and boot with synthetic and real workloads (Numbers, Pages, Safari) and produces a reproducible score against a baseline.
- Role
- Sole author: architecture, Swift app, backend and measurement methodology
- Outcome
- A fintech endpoint lab compares machines and system images using auditable raw data instead of a black-box number.
Commercial benchmarks like PCMark and Geekbench hand back a single score without exposing how each workload behaved under corporate agents. MacBenchmark was built for full certification transparency: before starting, a preflight check validates power, disk, network and running processes, while live telemetry tracks every stage.
To ensure reproducibility, workloads use deterministic seeds, discard warmup runs and execute serially with thermal cooldown pauses. The layered score is computed via weighted harmonic mean against a reference machine (baseline = 1000), with both the baseline and raw results exportable as JSON.
It also includes optional sync to a Node.js/TypeScript and PostgreSQL backend with an offline queue to consolidate lab runs. The code is private; happy to show it in a conversation.