Primate Labs, the Canadian developer behind PC benchmarking tool Geekbench 6, has unveiled a new addition to its benchmarking family: Geekbench AI 1.0.
Available across the big five desktop and mobile platforms, Geekbench AI allows you to benchmark your devices using a series of frameworks to determine their AI capabilities.
But why the need to test your own devices? The era of Artificial Intelligence is upon us, as tech companies fall over themselves in their attempts to convince us that computers can genuinely think for themselves. While the truth of AI – more a case of machine learning than true intelligence – doesn’t live up to the hype, that doesn’t mean the AI label is completely useless.
There are many examples of where machine learning is helping to transform people’s lives, from the big (robotic dentists anyone?) to the small (AI-powered image resizing and recolouring). Much of this work goes on in the cloud, but you’ll find a growing number of AI-powered tools can be installed on your computer or mobile devices to utilise their resources instead of some online server, sometimes even saving you money in the process.
Geekbench AI makes use of multiple AI frameworks capable of taxing your CPU and GPU, but what options you have depends on which platform you’re using. For example, Windows users can choose between ONNX (CPU-only) and OpenVINO (CPU and graphics) frameworks. If you choose OpenVINO, you then need to choose which backend (CPU or DirectML) to run, which determines which chip – CPU or GPU – gets tested. Then it’s simply a case of clicking or tapping ‘Run AI Benchmark’ and waiting for the tests to be run.
Once complete, Geekbench AI will display your results – within the mobile app or in your web browser if you’re benchmarking your computer. You’ll be shown a Geekbench AI Score, which comprises three elements – the key one is Quantized Score, which should be compared to Geekbench AI’s benchline: 1,500 based on an Intel Core i7-10700 machine. Scroll down to review a detailed list of results test-by-test.
Once done, you can then choose a different framework or backend and run a new benchmark. When it comes to interpreting these scores, the higher the better, but Geekbench offers a PDF download providing more context.
You can download Geekbench AI 1.0 for Windows, macOS, Linux, Android and iOS. It’s completely free on mobile, while desktop users can make do with the home edition or fork out $99 for a Pro license, which includes automated testing and offline results management among its extra features.
Your Comments & Opinion