Wondering how your current Linux system will handle AI and ML-related workloads? Well, now you can find out.
Primate Labs, makers of the hugely popular Geekbench software, has released a new preview build of their machine-learning benchmarking tool, Geekbench ML.
And it now supports Linux.
Geekbench ML launched on Android and iOS a few years back. But with Intel and AMD adding neural processing capabilities to their latest chips to power ML-accelerated workflows on desktop PCs and laptops, this app is branching out too.
The new Geekbench ML 0.6 preview the first to support Linux, Windows, and macOS.
While the latter 2 systems do get slick-looking GUI desktop apps, those of us Linux users have to make do with a command-line tool. Not a major drawback to be honest since it’s the stats that matter, not the window dressing.
The tool bundles in a number of machine learning workloads designed to model real-world AI use-cases. By running through these you’re able to find how well your CPU, GPU, and NPU (if one is present) handle “cutting-edge” computing tasks.
The included workloads include Depth Estimation (simulating portrait lens blurring features found in fancy photo editors), Style Transfer (copies the look of one image to another), and Image Super-Resolution (which upscales images by x4).
And since all models and data sets within the app are the same across all supported platforms, the Geekbench ML score you get on Linux can be reliably compared to the scores achieved on other platforms and devices.
Downloading Geekbench ML
As mentioned, Geekbench ML 0.6 is a preview release. You should expect quirks, inconsistencies, and bugs while running it. Primate Labs plan to issue a stable 1.0 release sometime next year, and feedback from the preview release will help inform and improve it.
You can download Geekbench ML from App Store for iOS, the Google Play Store for Android, and head to the Geekbench website download page to get Linux, macOS, and Windows builds.
The Linux version works on Ubuntu 18.04 LTS and above on Intel and AMD systems with 2GB RAM or greater. If your machine is modest, don’t expect expedient results!
Once done, you’re given a link to view your results online. From there you can compare to see how your hardware stacks up.
I’m sure whatever you’re using will be better than the laptop I tested ;)