Geekbench's new tool reveals how well your device handles AI
Geekbench, a renowned benchmarking utility, has introduced a new cross-platform tool named Geekbench AI. This innovative tool is designed to assess the performance of devices under heavy artificial intelligence (AI) workloads. It evaluates a device's CPU, GPU, and NPU (neural processing unit) capabilities in handling machine learning applications. The software was initially developed by Primate Labs under the name Geekbench ML and launched in preview in 2021 before being renamed to Geekbench AI.
Geekbench AI's unique performance evaluation approach
Geekbench AI adopts a unique approach to evaluate how different hardware responds to various AI-related tasks. It assesses performance based on both accuracy and speed, supporting multiple frameworks including ONNX, CoreML, TensorFlow Lite, and OpenVINO. The tool provides three scores: full precision, half precision, and quantized. According to Primate Labs, these scores also include an accuracy measurement to determine how close a workload's outputs align with the truth or "how accurately that model can do what it's supposed to do."
Geekbench AI's potential impact on device testing
The introduction of Geekbench AI could potentially change how devices with on-device AI are tested. For example, Copilot Plus PCs and new Pixel 9 phones may require more extensive testing to see how performance in real-world tasks correlates with Geekbench AI's numbers. The tool might be used for checking the accuracy of predictive text or what generative AI-enabled image editor comes up with, moving beyond traditional parameters like framerates or loading times.