PC Benchmark

What a browser benchmark can honestly tell you

This is worth saying plainly, because most online benchmarks do not. A web page cannot measure your hardware. It can only measure how fast this browser, on this machine, right now gets through a particular piece of work — and a great deal sits between that and your components.

• The browser's JavaScript engine is doing the work, and different browsers produce different numbers on the same machine.
• Other tabs and applications compete for the same cores. A background video call can halve a score.
• Thermal and power state matter. A laptop on battery, or one that has been working hard, throttles. The same machine can score forty per cent lower ten minutes later.
• The browser deliberately hides things. Core count is rounded, memory is capped at 8 GB regardless of what is installed, and graphics details are masked in some browsers.

So treat the score as a rough, relative figure: useful for comparing two browsers on one machine, or the same machine before and after a change. Not useful for comparing your computer to someone else's, and no substitute for a native benchmark.

Run it more than once

The first run is usually the slowest, because the JavaScript engine has not yet optimised the code it is about to run repeatedly. Run it three times and take the best.

What is included

• Processor — integer, floating point and string workloads, single-threaded.
• Memory — allocation and copy throughput.
• Graphics — WebGL shader fill rate.
• Storage — IndexedDB write and read throughput.

The combined score adds them together. It is an arbitrary number with no units, and it means nothing except in comparison with the same test run elsewhere. Do not compare it to any other benchmark's score.

The device panel

Above the test is what your browser will admit about the machine. Most of it is deliberately imprecise: core count is rounded, memory is capped at 8 GB, and the graphics name is masked in some browsers. It is shown here partly because it is useful and partly because the gap between it and reality is worth seeing.

How to use it

1. Close other tabs and applications first. This matters more than anything else you can control.
2. Plug the laptop in if it is on battery. Power saving changes the result substantially.
3. Press run and leave the tab in the foreground — browsers throttle background tabs.
4. Run it three times and use the best. The combined score moves by ten per cent or more between runs.

Common uses and tips

• A quick sanity check that a machine is performing as it did.
• Comparing browsers on one machine.
• Spotting thermal throttling by running it cold and again under load.
• Checking a second-hand machine quickly, before doing anything more thorough.
• Not for comparing two different computers. Too many variables sit between the score and the hardware.

Questions people ask

What does the combined score mean?

Nothing on its own. It is an arbitrary sum with no units, useful only for comparing the same test on the same machine or between browsers.

Why is it different every run?

Background activity, thermal state and JavaScript engine warm-up. Ten per cent variation between runs is normal.

Can I compare this to my friend's score?

Not meaningfully. Different browsers, thermal states and background activity make the comparison unreliable.

Does it test all my cores?

No. The processor test is single-threaded, so core count makes little difference.

Is anything installed or left behind?

No. The storage test deletes its database when it finishes, and nothing is sent anywhere.