Intel has announced that it is the first and only company to achieve full neural processing unit (NPU) support in the MLPerf Client v0.6 benchmark, marking a significant milestone in standardized AI performance evaluation on consumer PCs.
Full NPU Compliance: Intel’s Core Ultra Series 2 processors are the only ones to achieve complete NPU compliance in MLPerf Client v0.6, showcasing the company’s leadership in AI acceleration on client platforms.
Performance Metrics: The Core Ultra Series 2 processors demonstrated impressive performance, generating the first word in just 1.09 seconds (first token latency) and achieving a throughput of 18.55 tokens per second. This indicates rapid response times and efficient processing for AI tasks.
Benchmark Scope: MLPerf Client v0.6 evaluates performance across four content generation and summarization use cases based on the Llama 2 7B model.
Collaborative Development: The MLPerf Client benchmark is developed by the MLCommons consortium, which includes industry leaders like Intel, AMD, Microsoft, NVIDIA, and Qualcomm. The v0.6 release expands support to NPUs, reflecting the evolving landscape of AI hardware acceleration.
This achievement underscores Intel’s commitment to advancing AI capabilities on consumer devices, providing users with faster and more efficient AI-driven experiences.
Source : INTEL