How to Test Your Computer’s Artificial Intelligence Capabilities
Artificial intelligence, and generative AI in particular, is such a prominent technology now that it’s often mentioned first at product launches, from the Pixel 9 series to Microsoft’s Copilot+ line of PCs . These devices are getting more AI features and more components dedicated to AI tasks. If you’re wondering how well your PC can handle today’s most common AI tasks, check out the newly released Geekbench AI 1.0 from the benchmarking experts at Geekbench.
For those of you unfamiliar with benchmarking, it is essentially stress testing your hardware using specially designed tasks to help you get an idea of how powerful your system is and how it compares to others. You can test many different components of your computer, but in this situation we’re especially interested in AI performance testing.
What artificial intelligence tests actually test
Run an artificial intelligence test like Geekbench AI and it will test how well your computer can perform tasks that artificial intelligence typically performs: detecting objects and faces in images, creating a new image based on the style of an existing one, or translating between languages. You can use web apps for these tasks, but these tests will show you how well your computer can handle them locally.
These different tests use different types of AI, including machine learning (training models on huge amounts of data) and deep learning (a more sophisticated version of machine learning). They also measure various aspects of AI performance: for example, how quickly AI tasks can be processed, as well as how accurate they are.
Modern systems now often have dedicated neural processing units (NPUs), processors designed to handle artificial intelligence tasks. These tasks often require different types of calculations and workflows than more general computing processes, so having hardware components built specifically for them can speed things up (just like a GPU or GPU is designed specifically for visual effects processing).
Geekbench AI gives three scores at the end: full precision, half precision, and quantized. Full precision reflects AI tasks where a high level of precision is required and most of the demands are placed on the system, while the other two make slight compromises on accuracy and speed (often required if efficiency is important).
“Just as CPU-bound workloads vary in how they can take advantage of multiple cores or threads to scale performance (which requires both single-core and multi-core scores in most related benchmarks), AI workloads span a range of levels accuracy, depending on the task required, available hardware, and middleware platforms,” Geekbench’s John Poole explains in a blog post .
How to Run Geekbench AI on Your Computer
Head over tothe Geekbench AI download page to find packages for Windows, macOS, and Linux (as well as links to the Google Play Store and Apple App Store if you want to run tests on your mobile devices). All of these programs are easy to use, even if you have no experience with benchmarking software.
Geekbench AI’s main desktop interface primarily displays some information about your system, including the operating system you’re using and the processor you have installed. You then receive a description of what the test will cover. (Note that this will take a few minutes.)
You then get drop-down menus to configure the test. First is the AI Framework , which is essentially the set of AI tools you want to use for testing: the ones you see will depend on your system and may include ONNX (Open Source Neural Network Exchange ) and Core ML (an artificial intelligence framework used on Apple hardware).
There is also AI Backend , that is, a system component whose capabilities you want to test. This will depend on what’s inside your computer, but you’ll see the CPU and possibly also the GPU and NPU (or Neural Engine ). On some systems, you’ll also see an “AI Device” option that allows you to switch between available processors.
Once you’re happy with how the test is set up, click “Run AI Test” to run it. Your system’s results will appear in your web browser, and you can view other results posted from other devices through the same portal. Note that scores are calibrated to a base level of 1500, which reflects the performance of the Intel Core i7-10700 processor. Higher scores indicate better AI management.