Announcing Geekbench AI: Measure AI Performance on Devices

MetaversePlanet August 19, 2024Last Updated: August 19, 2024

0 1 minute read

Geekbench has announced its new tool, “Geekbench AI,” designed to measure the performance of artificial intelligence on various devices. This cross-platform AI benchmarking tool uses real-world AI tasks to evaluate how well your phone or computer can handle AI features. By assessing the performance of the CPU, GPU, and NPU, Geekbench AI determines whether your device is equipped for both current and future advanced AI capabilities.

Geekbench AI provides a comprehensive evaluation of AI performance by running ten different AI workloads across three different data types. With large datasets, both developers and smartphone users can assess the AI performance on their devices in just a few minutes, receiving scores for single-precision, semi-precision, and quantized calculations.

You can see your device’s AI performance at the hardware level

Geekbench AI measures AI performance at the hardware layer in detail, allowing you to test the GPU, CPU, or custom NPU on your device. You can also select from available AI frameworks, such as Core ML or QNN, to find the optimal combination. This feature enables developers to identify the most suitable frameworks and models for specific workloads, while end users can easily measure the impact of custom AI hardware.

Geekbench AI allows you to compare performance across platforms like Android, iOS, Windows, macOS, and Linux by running the same workloads. Whether you’re testing a smartphone with an ultra-low-power NPU or a custom workstation with kilowatts of AI computing power, you can instantly compare results using the Geekbench AI results scanner.

For professionals, Geekbench AI offers a package that includes automated testing tools, offline mode, and a commercial use license. Organizations can also obtain site and resource licenses, which provide dedicated support, updates, and access to source code.To download Geekbench AI Here You can click on the link.