UL Solutions launched the new version of Procyon AI Inference Benchmark, an artificial intelligence inference performance measurement tool, this year. It is now also available for the macOS operating system, mainly adding support for the Core ML AI inference engine to measure the artificial intelligence computing performance of Mac models equipped with Apple Silicon processors.
Leveraging the Core ML AI inference engine, the Procyon AI Inference Benchmark runs a computer vision benchmark on macOS, following the same AI inference workload as the Windows operating system. It tests the performance of four different combinations: CPU, CPU+GPU, CPU+AI computing element, and CPU+GPU+AI computing element.
All test workloads are designed to meet the hardware requirements of current Mac models, including fully compatible GPUs and AI components. In addition to measuring raw hardware performance, the measurement process also considers the accuracy of AI inference, comparing FP16 (half-precision floating-point) and INT precision inference results, making it easier to assess specific performance.
On the other hand, the new version of Procyon AI Inference Benchmark already supports artificial intelligence inference engines including Microsoft WindowsML, Qualcomm SNPE, Intel OpenVINO, and NVIDIA TensorRT, and is compatible with Intel's new 14th-generation Core series laptop processors code-named Meteor Lake with NPU design, or Qualcomm's new processors for Windows laptops.
The Procyon AI Inference Benchmark is now open for application. UL Solutions also stated that many companies have begun using this tool to measure the AI computing performance of laptop devices.
