Geekbench launches a new cross-platform AI performance testing tool Geekbench AI, which is used to evaluate the CPU, GPU and NPU performance of devices when processing machine learning tasks. This tool is improved based on the Geekbench ML preview version previously released by Primate Labs, and supports multiple frameworks such as ONNX, CoreML, TensorFlow Lite, and OpenVINO. It provides three scoring standards of full precision, half precision, and quantification, and also includes accuracy measurements. to more fully evaluate equipment performance.
Geekbench's development team, Primate Labs, originally developed the software under the name Geekbench ML and released a preview version in 2021. As technology develops and needs change, the software is renamed Geekbench AI, aiming to explore the performance of different hardware in dealing with various tasks. The tool evaluates performance based on accuracy and speed and supports multiple frameworks, including ONNX, CoreML, TensorFlow Lite, and OpenVINO.
Geekbench AI provides three scoring criteria: full precision, half precision and quantification. Primate Labs says the scores also include accuracy measurements, which evaluate how closely a workload's output matches reality, i.e. how accurately the device performs the task.
More time will be needed to get a fuller picture of how a device performs in actual tasks in relation to Geekbench AI's scores. For example, for devices with built-in intelligence, we might not just look at frame rate or load times, but also the accuracy of predictive text or the quality of the results generated by an image editor.
Currently, the tool can be downloaded on Windows, macOS, Linux, Android and iOS platforms, and users can try and experience its functions on their own.
The launch of Geekbench AI provides developers and users with a standard to objectively evaluate the AI performance of devices, helping to promote the further development of AI hardware and software. Its cross-platform feature also facilitates the use by users of different systems, and it will play an important role in the field of AI performance testing in the future.