Browser benchmarks vs app benchmarks: what each can measure

Installed benchmark apps reach parts of a device a web page cannot, and a web page runs where no app can be installed. What each kind of test measures, where browser results are weaker, and when to use which.

Benchmark apps you install — Geekbench, 3DMark, AnTuTu and others — and tests that run in a web page, like DeadTier, can both tell you how fast a device is. They are not interchangeable. Each can see things the other cannot, and knowing the difference tells you how much weight to put on a result.

What an installed app can do

Run native code. An app runs code built for the processor directly, with fewer layers in between, so its numbers reflect the hardware more closely.
Use the graphics chip properly. Apps talk to the graphics hardware through low-level interfaces such as Vulkan, Metal and DirectX, and can run long, demanding 3D scenes. That is how a real graphics benchmark works.
Read sensors. Depending on the platform, an app may be allowed to read things like temperatures and clock speeds, which a web page cannot.
Run longer and broader tests. Many apps run several kinds of work over several minutes and compare the result with large databases of other devices.

What a browser test can do that an app cannot

Run without installing anything. That is the whole point of DeadTier. You can test a demo unit in a shop, a second-hand phone before you hand over any money, or a work laptop where you are not allowed to install software.
Work the same way everywhere. One page runs on Android, iPhone, Windows, macOS and Linux, so a phone and a laptop take exactly the same test.
Take about a minute. There is no download, no account and no permission to grant.
Stay out of your data. A web page cannot read your files, contacts or other apps, and DeadTier stores its measurements without anything that identifies you or your device.

Where a browser test is weaker

It runs inside the browser. JavaScript is fast today, but it adds layers between the test and the hardware, and two browsers on the same device can produce different results.
It does not choose its cores. The test starts one worker per core, but the operating system decides where each one runs. On chips with fast and slow cores, that adds variation from one run to the next.
It cannot see temperatures or clock speeds. There is no web API for processor temperature, so DeadTier never shows one. It can see the effect of heat — speed falling during the run — but not the heat itself.
It cannot honestly benchmark a graphics chip. DeadTier’s drawing test is 2D canvas work that leans on the processor, which is why it is called a render test. For graphics performance, use an app built for it.
Background tabs are slowed down. Browsers deliberately slow pages you are not looking at, so a browser test only works while it stays in front.

Why the numbers do not match

Every benchmark has its own workload and its own scale, so a Geekbench result, an AnTuTu result and a DeadTier score cannot be converted into one another. What should agree is the order: if one device is clearly faster than another in an app, it should usually be faster here too. If two tests disagree about the order, check the conditions first — the charger, heat and background apps move every benchmark.

When to use which

Use an installed app when you own the device, want the deepest picture or care specifically about graphics. Use a browser test when you cannot or would rather not install anything, when you want to compare very different kinds of device on one scale, or when you need an answer in a minute — in a shop, at a seller’s door, or on someone else’s laptop. The two answer different questions well, and a careful buyer can use both.

If you only want to know which graphics chip a device has, rather than how fast it is, what GPU do I have? reads the name the browser reports, without measuring anything.

Measure the device you are holding

About a minute, nothing installed, and it works on a device you do not own.

Run the test