Benchmarking Browsers
summary
A cross‑vendor benchmark called Speedometer 3 defines concrete workloads (updated 2do MVC, charts, editors, news-like page navigation) and a joint governance model across Gecko WebKit and Blink. Engineers run the suite, profile hotspots, fix issues (a canvas bottleneck drops from 760 to 144 samples, ~80% reduction) and measure user impact: Firefox shows ~4% score gain, median first contentful paint improves ~45–50 ms, 95th‑percentile JS time drops ~300 ms and 95th key‑press latency falls ~6 ms.
AI-written summary · about
chapters (23)
- 0:00Browser benchmarking intro
- 1:29Why browser performance matters
- 3:38History of benchmark tools
- 7:57Challenges with traditional benchmarks
- 10:06Why RUM isn't enough
- 11:11Lab-based benchmarks advantages
- 14:02Speedometer 3 project evolution
- 16:08Speedometer 3 goals
- 18:37Speedometer 3 design principles
- 20:02How Speedometer 3 measures perf
- 22:10Choosing representative workloads
- 25:06Modernizing Todo MVC workload
- 27:57Adding chart workloads
- 29:24Including editor workloads
- 30:28Simulating static page loads
- 31:50Optimizing Firefox with Speedometer 3
- 36:07Validating benchmark improvements
- 41:31Speedometer future & community
- 42:56Q&A: Biggest performance wins
- 44:44Q&A: Energy vs performance
- 46:31Q&A: Frameworks vs browser priorities
- 48:14Q&A: Speedometer 3 optimizations
- 49:40Q&A: Contributing to Speedometer
