AI Hardware Performance Rankings

Welcome to our AI Performance Rankings page. Here, we evaluate the components based on their AI processing power, measured in TOPS (Tera Operations Per Second) – a critical metric indicating the computational throughput, particularly for AI tasks.

The first column shows peak performance for INT8/FP8 precision, which is the most widespread metric for evaluating AI inference capabilities. We exclude Sparsity to provide a more accurate reflection of AI performance in dense computation scenarios where sparsity optimizations may not be applicable. The second and third columns show the performance with Sparsity, and FP4 TFLOPS, when supported.

For SoCs, the results reflect the peak performance of the integrated NPU. Additionally, it’s important to note that, according to Microsoft, a NPU must have at least 40 TOPS of AI computing power for the PC to be considered “AI-capable.”