What GPU do I have?

The graphics chip your browser reports, and the family it belongs to. Read from the device you are holding - nothing installed, and nothing measured.

• Reported by the driverreading...
• Logical CPU coresreading...
• Reported memoryreading...

All three are what the browser says, not what the box says. The core count is logical cores and tells you nothing about whether they are the same size as each other - most phones mix fast and slow ones. The memory figure is deliberately coarse, and plenty of browsers report neither.

This is a name, not a score

Nothing above was measured. The browser was asked which graphics chip it is drawing with, it answered with a string, and that string was matched against published GPU families. You could look the same name up yourself, which is exactly what makes it worth showing - it is checkable.

No web page can honestly measure how fast a graphics chip is. There is no API that gives a browser tab the kind of access a native benchmarking app has, so anything on the web claiming to put a number on your graphics hardware is measuring something else and naming it after the chip. This site runs a drawing workload of its own and calls it a render test for exactly that reason: it is substantially processor-bound, so it is not named after a part it cannot isolate.

Why it often says nothing useful

A blank or generic answer above is common and correct, and it is not a fault on your device:

• Safari says "Apple GPU" and stops. Every iPhone, iPad and Mac gets the same string, from the oldest to the newest, so it identifies nothing.
• Some browsers withhold the name on purpose. A precise driver string is unusual enough to help identify a visitor across sites, so hiding it is a privacy decision rather than a bug.
• Some chips are not in the table. When a name matches no known family it is printed exactly as reported and nothing is claimed about it. A guess would be worth less than the blank.

What the family name is good for

On its own, not much - it tells you roughly what class of part the maker put in. It becomes useful when you put it beside a measurement of the same device, because then the two can disagree. A chip from a strong family that measures like a weak one is a device performing below what it should, and that gap is the interesting finding. It is never proof of anything dishonest: heat, a background app and an aged battery all produce exactly the same result.

The gap points both ways, which is the part usually left out. If a chip from a weak family measures like a strong one, the likeliest explanation is not a surprisingly fast phone - it is that the scale doing the measuring is too generous. The longer version of all this covers where the names come from and what each family means.