Geekbench 7 massive CPU testing overhaul with AI and CUDA support

Primate Labs has just dropped the latest iteration of its popular cross-platform benchmark, Geekbench 7, bringing a suite of significant updates designed to deliver a far more realistic picture of how modern CPUs and GPUs actually perform in real-world scenarios.

Forget forcing every workload to use every single core. One of the biggest shifts in Geekbench 7 is how multi-core performance is measured. The new system now mirrors actual application behavior. For instance, web browsers are typically single or lightly threaded, so tests like the HTML5 browser test have been excluded from the multi-core suite unless the real application is actively utilizing multiple threads. This refinement ensures that the benchmark provides a much more honest and accurate indication of overall system performance.

To truly evaluate multimedia processing, Geekbench 7 introduces powerful new media workloads. These tests delve into how CPUs handle complex daily tasks, including encoding AV1 video for screen sharing during calls, compressing audio using the Opus codec for voice recorders, and simultaneously decoding audio and video while generating live captions using OpenAI’s Whisper speech recognition model. This means system performance is now measured against demanding content streaming and conferencing tasks.

The visual processing capabilities of Geekbench 7 have also been dramatically enhanced. The GPU benchmarks are no longer confined to gaming; they now heavily emphasize machine learning and content creation. New tests assess critical tasks like real-time face tracking, applying filter effects for social media, AI image upscaling, and background blurring during video calls.

Furthermore, the benchmark has expanded into professional creative fields by introducing workloads for RAW image processing, LUT-based video color grading, path tracing, and fluid simulation. This allows developers and enthusiasts to gauge GPU performance in demanding visual effects scenarios with unprecedented detail.

For gaming enthusiasts, a new Game Physics benchmark based on the Jolt Physics engine has been added, providing a more accurate measurement of CPU capabilities specifically for physics-heavy gaming environments. The Photo Editor workload has also received an upgrade, featuring more complex editing operations and expanded support for advanced image formats like JPEG XL and DNG within the Photo Library test.

In terms of foundational performance, Geekbench 7 updates its datasets to reflect the larger and more complex files users handle today. The File Compression test now incorporates a broader mix of archives containing source code, object code, and text documents, while the PDF Viewer workload processes a wider range of file types, including technical documentation and academic papers.

Finally, Geekbench 7 is making it easier for Nvidia users to benchmark their powerful GPUs. The benchmark now includes support for Nvidia’s CUDA API, alongside OpenCL, Vulkan, and Metal, allowing users to assess GPU performance using the same APIs prevalent in professional and AI applications.