Geekbench: New AI test for iOS, iPadOS and macOS, Windows and Android

It’s been in testing since last winter, and now it’s officially ready: Geekbench put a new suite of benchmarks for artificial intelligence skills online this week. The benchmark was originally called Geekbench ML (ML stands for machine learning), now called Geekbench AI (for artificial intelligence). In version 1.0, as in previous Geekbench versions, Apple devices, Android and Windows can be tested – at Apple these are currently Mac, iPad and iPhone.

Advertisement

The first ranking is out

even one First ranking has already been published. Geekbench divides its AI tests into three areas: single precision, half precision and quantized. In single-precision and half-precision values, the RTX 4090 series graphics cards are the leader, while in the quantized result it is the iPads with the M4 SoC. Additional data such as processor frequency, technology, cores and RAM can be accessed for each device listed. In some cases there also seem to be error messages; many Macs and other devices are listed with only one point in all categories.

Geekbench is still publishing Results released from beta phase with Geekbench ML – Here the values ​​are also separated according to CPU, GPU and NPU (for Apple: Neural Engine), which is what you would also want with Geekbench AI. However, it is currently not possible to download Geekbench ML, you will be redirected to Geekbench AI.

Apple in India: spectacular growth, iPhone 16 Pro production begins

Real-world AI results

Geekbench promises that Geekbench AI can test the “real-world performance” of devices in the AI ​​​​field. “Geekbench AI is a cross-platform AI benchmark that uses real-world machine learning tasks to evaluate the performance of AI workloads. Geekbench AI measures your CPU, GPU, and NPU to determine if your device is ready for today’s and future cutting-edge applications about machine learning,” the creators wrote.

In addition to the standard free version, Geekbench also offers one priced at $99 Commercial version of AI Benchmark. It allows use in standalone mode, can process results both online and offline, and offers an automation mode.

(B.Sc.)