Screen Test Pro

Leaderboard

The best runs of the browser benchmark as they arrive: one row per machine, up to 200, from anonymous samples. The homepage shows the top ten of the last 30 days; this is the rest of it, for any version we still hold, over 30 days or all time.

60–90 s · fullscreen · run it with sharing on and your machine gets a row

Top 200

loading…
#ScoreTierMachineDay

How to read this

Every row is one shared run of the Standard profile that was fully measured — complete, or thermally throttled and marked as such. One row per machine: the query keeps a device's best run and drops its repeats, using only the coarse fields we store (GPU, OS and browser family, browser major version, core and memory buckets, form factor, device family), so no identifier is involved and none is needed. Degraded and compatibility runs never rank; their scores are floors, and ranking floors against peaks would mislead everyone.

Versions never mix. Benchmark 2.2.0 is the current workload; earlier versions stay selectable for as long as we hold their samples (24 months), because a score only means something next to scores from the same workload. If you want what a piece of hardware typically scores rather than its best day, the hardware benchmarks give medians with quartiles over groups of at least thirty samples. What each row contains, and what it never does, is spelled out on the privacy page.

Share this page
DiscussionOne thread per language, shared across the whole site. Sign in with Google to post.
Discussion