Geekbench ML has long been the standard for benchmarking processors’ AI capabilities, but this is set to change with the release of Geekbench AI by Primate Labs. This new tool is designed to provide a more accurate assessment of a processor’s performance with AI workloads.
Like its predecessor, Geekbench AI allows users to evaluate the CPU, GPU, and NPU across various AI tasks. However, it now supports several new frameworks that were not previously available.
The latest Geekbench AI app now includes support for Qualcomm QNN on Windows, allowing you to benchmark the Snapdragon X Elite’s NPU in addition to its CPU and GPU for AI workloads.
In addition, the app has expanded its support to include the OpenVINO framework (widely used by Intel), TensorFlow Lite, ONNX, CoreML, Samsung ENN, ArmNN, and Qualcomm QNN on Android, among others. Geekbench AI also now provides three distinct types of scores: Single Precision, Half Precision, and Quantized.
The Single Precision score reflects accuracy in performing AI tasks, while the Half Precision score is typically higher due to its lower precision. The Quantized score, on the other hand, uses the INT8 data type, which is a low-precision format.
The new Geekbench AI app is available for download and can be used immediately on Windows, macOS, Linux, Android, and iOS.
0 Comments