Geekbench 7 launches with overhauled tests for AI workloads, CUDA support, and real-world benchmarking

Geekbench 7 introduces biggest overhaul yet -- real-world CPU testing, new media workloads, AI benchmarks, and CUDA support

Primate Labs has released Geekbench 7, the latest version of its popular cross-platform benchmark. The update introduces new and improved workloads designed to better represent and measure the performance of the latest CPUs and GPUs across major platforms, including Android, iOS, Windows, macOS, and Linux. Like previous versions, Geekbench 7 is available free of charge for personal use. Enthusiasts and IT professionals can also opt for Geekbench 7 Pro, which is available at a 20% discount until August 6. One of the biggest changes introduced with Geekbench 7 is how the multi-core benchmark behaves. It will now mirror real-world application behavior instead of forcing every workload to use all available CPU cores. For instance, the HTML5 browser test has been excluded from the multi-core suite since web browsers are generally single- or lightly threaded. Geekbench 7 will only run a benchmark in multi-threaded mode if the real application it's modeling also uses multiple threads. As a result, Primate Labs says the new benchmark will deliver a more realistic indication of overall system performance. Geekbench 7 will also feature a new set of media workloads designed to better evaluate how modern CPUs handle everyday multimedia tasks. These include encoding AV1 video for screen sharing during video calls, compressing audio using the Opus codec for applications such as voice recorders and podcasts, and decoding audio and video while simultaneously generating live captions using OpenAI's Whisper speech recognition model. Essentially, these workloads are claimed to better represent tasks such as video conferencing, content streaming, and media consumption. A new Game Physics benchmark is also being added, which is based on the Jolt Physics engine used in modern games, to offer a more accurate measurement of CPU performance in physics-heavy gaming scenarios. The update further enhances its Photo Editor workload with more complex editing operations and expands the Photo Library test to include support for newer image formats such as JPEG XL and DNG. The GPU benchmark has also been revamped to go beyond gaming, with greater emphasis on machine learning and content creation workloads. The new tests will measure tasks such as real-time face tracking and filter effects for social media apps, image upscaling using AI, and background blurring during video conferencing. Additionally, the update adds new GPU workloads for RAW image processing, LUT-based video color grading, path tracing, and fluid simulation. Geekbench 7 is also set to introduce support for Nvidia's CUDA API in addition to OpenCL, Vulkan, and Metal, allowing users to benchmark Nvidia GPUs using the same API that is used in a variety of professional and AI applications. Lastly, in order to keep up with demanding modern workloads, the benchmark datasets have been updated to better reflect the larger and more complex files that users work with today. For instance, the File Compression test now includes a broader mix of archives containing source code, object code, and text documents, while the PDF Viewer workload processes a wider range of files, from park maps to technical documentation and academic papers. The update also introduces more assets and additional image formats across its developer and image processing workloads. Follow Tom's Hardware on Google News, or add us as a preferred source, to get our latest news, analysis, & reviews in your feeds.

Geekbench 7 is out with overhauled performance tests for CPU and GPU workloads

Primate Labs released Geekbench 7 today, promising benchmarks that more closely reflect how people use their devices across everyday tasks, gaming, and AI workloads. Here's what's new. Geekbench 7 now available When Primate Labs released Geekbench 6 in early 2023, the company said that "a lot [had] changed in the tech world in the past three years," with AI and ML becoming "ubiquitous in general and mobile applications." Three years later, that is even more true. From on-device AI inference and image processing to video editing and increasingly demanding games, CPU and GPU performance now matter across a wider range of everyday tasks. Today, Primate Labs released Geekbench 7, featuring "both new and improved workloads to measure the performance of your CPUs and GPUs." Geekbench 7 also revamps the app's GPU Benchmark, placing greater emphasis on machine learning and content creation workloads. Here's Primate Labs: For machine learning, the GPU benchmark includes workloads that: Geekbench 7 also introduces new GPU image editing and synthesis workloads, including RAW image processing, LUT-based video color grading, path tracing, and fluid simulation. Additionally, Geekbench 7 adds Nvidia's CUDA to its list of supported GPU APIs, alongside OpenCL, Vulkan, and Metal. Another improvement over Geekbench 6 is how the updated benchmark handles multi-core performance. Primate Labs explains that Geekbench 7 now runs workloads across multiple threads only when the real-world task being tested typically does the same. For example, the HTML5 Browser test is no longer included in the multi-core suite because web browsers are typically single-threaded or only lightly threaded. When it comes to media performance, Geekbench 7 adds new workloads for measuring how CPUs handle audio and video encoding, decoding, and processing: * Encode screen-sharing video with the AV1 codec, modeling the screen-sharing features in video conferencing apps. * Compress music and spoken-word audio with the Opus codec, modeling voice memo and podcast apps. * Decode video and audio while generating live captions with the Whisper speech recognition model, modeling video playback with automatic subtitles enabled. Finally, Geekbench 7 also increases the size and variety of the data sets used throughout its tests. The File Compression workload, for instance, now includes more archives spanning source code, object code, and text documents. The PDF Viewer test, in turn, covers a broader range of files, including park maps, technical documents, and academic papers. Likewise, developer and image-processing workloads have also been updated with more assets and additional image formats. To celebrate the launch of Geekbench 7, Primate Labs is offering 20% off Geekbench 7 Pro through August 6. The standard version remains free for personal use. To learn more about Geekbench 7, follow this link. Worth checking out on Amazon