The favored benchmarking utility Geekbench has launched a brand new cross-platform software to judge the efficiency of gadgets underneath AI-heavy workloads. Geekbench AI measures a tool’s CPU, GPU, and NPU (neural processing unit) to find out how nicely it could deal with machine studying purposes.
Geekbench developer Primate Labs has been engaged on the software program utilizing the identify Geekbench ML, which launched in preview in 2021 however shifted the identify to AI for causes that appear apparent. To discover how totally different {hardware} responds to totally different AI-related duties, it evaluates efficiency based mostly on each accuracy and velocity, with help for various frameworks, together with ONNX, CoreML, TensorFlow Lite, and OpenVINO.
It delivers three scores, full precision, half precision, and quantized. Primate Labs says the scores even have an accuracy measurement to judge how shut a workload’s outputs are to the reality, “or how precisely that mannequin can do what it’s alleged to do.”
We’ll want extra time with gadgets operating on-device AI resembling Copilot Plus PCs and the entire new telephones to see how efficiency in real-world duties correlates with Geekbench AI’s numbers. Checking framerates or loading instances is one factor — now we is perhaps checking the accuracy of predictive textual content, or what generative AI-enabled picture editor comes up with.
You may obtain the software now to strive it your self on Home windows, macOS, Linux, Android, and iOS.