Exploring the Artificial Intelligence Potential of Your Personal Computer
artificial intelligence and generative AI have become so prevalent, they're frequently mentioned prior to any other tech at product launches - from the Pixel 9 series to the Microsoft Copilot+ PC collection. These devices are incorporating more AI features and components dedicated to AI tasks. If you're curious about your computer's AI capabilities, consider the recently released Geekbench AI 1.0.
For those unacquainted with benchmarking, this process involves putting your hardware through its paces with specialized tasks, giving you an idea of its power and comparing it to other systems. In this instance, we're focusing on AI performance testing.
What AI benchmarks actually test
Running an AI benchmark such as Geekbench AI will evaluate your system's capability to undertake tasks that AI typically handles - like identifying objects and faces in photos, generating a new image based on an existing style, or translating between languages. You might use web applications for these responsibilities, but these benchmarks gauge how well your computer can manage these tasks locally.
These diverse tests utilize a range of AI technology, including machine learning (training models on massive data sets) and deep learning (a more intricate form of machine learning). They also assess different facets of AI performance, such as how swift AI tasks can be processed and their accuracy.
Today's systems frequently include Dedicated Neural Processing Units (NPUs), processors designed for handling AI tasks. These tasks necessitate distinct calculations and workflows compared to general computing processes, thereby enabling hardware components specifically built for them to expedite everything (similar to how a GPU or Graphics Processing Unit accelerates visuals).
Geekbench AI generates three scores at the conclusion: Full Precision, Half Precision, and Quantized. Full Precision reflects AI tasks that demand high accuracy and impose major system demands, while the other two compromise on precision for speed (often required for efficiency).
"Just as CPU-bound workloads vary in their ability to leverage multiple cores or threads for performance scaling (hence, the need for both single-core and multi-core metrics in most related benchmarks), AI workloads span a spectrum of precision levels, depending on the task at hand, available hardware, and intervening frameworks," explains Geekbench's John Poole in a blog post.
Geekbench AI will give you three final scores.Credit: Lifehacker
How to run Geekbench AI on your computer
Navigate to the Geekbench AI download page to find the software packages for Windows, macOS, and Linux (along with links to the Google Play Store and Apple App Store for mobile device benchmarking). All these applications are user-friendly, even if you have no prior experience using benchmarking software.
The primary Geekbench AI interface on desktop displays information about your system initially, including the operating system and installed processor. Subsequently, it provides a description of the benchmark details. (Note that it takes several minutes to complete.)
Next, you'll find drop-down menus to configure the benchmark. Initially, you'll encounter the AI Framework, which represents the AI software suite for the test: The options available will vary according to your system, possibly including ONNX (the open-source Open Neural Network Exchange) and Core ML (the AI framework employed on Apple hardware).
Then comes the AI Backend, indicating the system component you want to test for AI capabilities: This depends on your computer's constituents, yet you'll normally see CPU, and possibly GPU and NPU (or Neural Engine) too. On certain systems, you'll also see an AI Device option, which allows you to switch between available processors.
Once you're satisfied with the benchmark configuration, click Run AI Benchmark to initiate it. Upon completion, your system's results will display in a web browser, enabling you to compare them to others' results via the same portal. Note that scores are calibrated against a baseline of 1,500, which reflects the performance of an Intel Core i7-10700 processor. Higher scores signify more proficient AI handling.
To test your computer's AI capabilities, consider using the recently released Geekbench AI 1.0. This benchmark evaluates your system's performance in tasks commonly handled by AI, such as object and face recognition, image generation, and language translation.
Geekbench AI's test utilizes various AI technologies, including machine learning and deep learning, to assess the speed and accuracy of AI tasks. The benchmark generates three scores: Full Precision, Half Precision, and Quantized, each reflecting different levels of precision and system demands.
