Popular benchmarking utility Geekbench has launched a new cross-platform tool for evaluating device performance under artificial intelligence-intensive workloads. Geekbench AI measures a device’s CPU, GPU, and NPU (Neural Processing Unit) to determine its ability to handle machine learning applications.
Geekbench developer Primate Labs has been developing software called Geekbench ML, which launched in preview in 2021, but changed the name to AI for obvious reasons. To explore how different hardware responds to different AI-related tasks, it evaluates performance in terms of accuracy and speed, and supports different frameworks, including ONNX, CoreML, TensorFlow Lite, and OpenVINO.
It provides three fractions: full precision, half precision, and quantized. Primate Labs says the scores also feature an accuracy measure, which evaluates how close a workload’s output is to the real world, “or how accurately the model does what it’s supposed to do.”
We need more time to study devices running on-device AI, such as Copilot Plus PCs and all new phones, to understand the relationship between performance in real-world tasks and Geekbench AI data. It’s one thing to check frame rates or load times – now we might be checking the accuracy of predictive text, or the capabilities of an AI-enabled image editor.
You can download the tool now and try it out for yourself on Windows, macOS, Linux, Android, and iOS.
