UL already has an AI Inference benchmark, which is based on an AI Computer Vision workload and is meant to measure and compare AI performance on lightweight PCs, but the new AI Image Generation benchmark, based on the Stable Diffusion AI model, is meant to measure performance on modern discrete GPUs and high-end hardware.

As said by UL, the performance range of consumer AI-capable hardware has become incredibly broad, thus, measuring AI Inference performance now requires a selection of benchmarks in order to optimally measure that same AI performance.

UL adds that the new AI Image Generation Benchmark is considerably heavier than the computer vision benchmark, and includes two tests built using different versions of the Stable Diffusion model, with more tests coming in the future. Currently, the AI Image Generation Benchmark supports Intel OpenVINO, NVIDIA TensorRT, and ONNX runtime with DirectML, with more inference engines planned for the future.

The UL UL Procyon benchmark suite offers flexible licensing, and you can check out more information over at the UL Procyon benchmark website.