2026-09-28T11:19:32.499219Z

Apple's A20 Pro shatters Geekbench 7 single-core record — 2nm chip beats desktop Intel Core i9 and AMD Ryzen 9 by up to 32%

www.tomshardware.com/pc-components/cpus/apples-a20-pro-shatters-geekbench-7-single-core-record-2nm-chip-beats-desktop-intel-core-i9-and-amd-ryzen-9-by-up-to-32-percent

Tom's HardwareApple's A20 Pro shatters Geekbench 7 single-core record — 2nm chip beats desktop Intel Core i9 and AMD Ryzen 9 by up to 32%Eats Intel's entry Panther Lake CPUs for breakfast too.

2026-09-29T02:25:16.929589Z

Auli@Auli@lemmy.ca

⁨17⁩d

Replying to @⁨KoalaUnknown@lemmy.world⁩

And can do that for two seconds before throttling.

2026-09-29T06:15:00.482479Z

tal@tal@lemmy.today

⁨17⁩d

Replying to @⁨Auli@lemmy.ca⁩

I’m not familiar with GeekBench, but I assume that, as with benchmarking software that I’ve used in the past, its benchmark runs for more than two seconds to help make numbers reproducible.

⁨Sep⁩ ⁨29⁩, ⁨2026⁩, ⁨06:15⁩Edited ⁨⁨Sep⁩ ⁨29⁩, ⁨2026⁩, ⁨23:18⁩⁩en

2026-09-30T06:15:44.728578Z

Kissaki@Kissaki@feddit.org

edited⁨16⁩d

Replying to @⁨tal@lemmy.today⁩

I haven’t checked what they do either, but I don’t see that as a reasonable assumption.

If they test for short burst performance specifically then the short runtime is the point. You then use multiple iterations of the test run to get a reasonable, reproducible result by averaging.