Geekbench 7 adds AV1 encoding and AI video upscaling benchmarks

Geekbench 7 formalizes AI and AV1 as standard performance metrics, shifting the industry benchmark from raw synthetic power to specific streaming and ML workloads. For streaming engineers, this provides a standardized way to evaluate silicon efficiency for background blurring, real-time upscaling, and low-latency codecs. By excluding single-threaded tasks like web browsing from multi-core scores, the update forces chipmakers to optimize for sustained media throughput rather than just high core counts. Watch for how server-side CPU manufacturers like Ampere or AWS Graviton adopt these benchmarks to market their AV1 encoding efficiency.

The release of Geekbench 7 occurs as the hardware landscape shifts toward NPU-integrated silicon (Neural Processing Units) becoming standard in consumer and server hardware. Per Notebookcheck (July 2026), preliminary testing under the new scoring system shows a shift in the performance hierarchy; for example, the AMD Ryzen 7 7700 now serves as the 2,500-point baseline, replacing the Intel Core i7-12700 used in previous versions. This recalibration is necessary as modern OS features, such as Apple’s iPadOS 27, increasingly rely on dedicated ML accelerators for background tasks like video transcription and real-time object extraction.

Furthermore, the inclusion of CUDA support marks a return of Nvidia-specific optimization within the benchmark suite. As reported by 9to5Mac (July 2026), the GPU benchmark now places significantly more weight on content creation and ML-driven effects than on traditional 3D graphics rendering. This aligns with broader industry trends where GPU compute is utilized for video color grading via Look-Up Tables (LUTs) and path tracing in virtual production. Primate Labs has also increased the size and variety of datasets for file compression and image processing to reflect the larger assets handled by modern professional media pipelines including JPEG XL and DNG formats.