Geekbench developer PrimateLabs has released a new preview version 0.6 of GeekbenchML, which extends beyond iOS and Android to provide benchmarking capabilities for artificial intelligence and other machine learning workflows on desktop platforms such as macOS, Windows and Linux.
AI and ML related workflows are not limited to mobile platforms, the hardware architecture of desktop and laptop devices is also changing to accommodate this shift in computing.
In the latest 0.6 preview, GeekbenchML now supports Windows, macOS, and Linux. This means you can see how machine learning-driven tasks run on your desktop, laptop or even server, regardless of whether your desktop, laptop or even server has new AI-specific hardware. And, as always, the models and datasets are the same across all supported platforms, so scores are comparable.
With a new and improved framework, GeekbenchML supports the use of newer models on the latest hardware, including CoreML on iOS and macOS, resulting in faster performance.
New workloads for GeekbenchML include: depth estimation, which reflects capabilities associated with software-assisted portrait mode photo effects; style transfer, which measures the ability to transform one image into the style of another; and image super-resolution, which evaluates AI-driven image upscaling.
GeekbenchML 0.6 is still a preview version and PrimateLabs will continue to adjust it based on test results and feedback, so scores from version 0.6 cannot be directly compared to scores from the previous version 0.5. PrimateLabs hopes to release a complete version 1.0 sometime in 2024.
GeekbenchML is already available on the App Store and Google Play Store, and PrimateLabs also provides preview versions of macOS, Windows and Linux directly on its website.