Introduction to MLPerf Inference V4.1 Results
Recently, MLCommons published the results of its industry-standard AI performance benchmark suite, MLPerf Inference V4.1. These results highlighted Intel’s achievements in AI performance, especially with their 5th Gen Intel Xeon Scalable processors and the new Intel Xeon 6 processors with performance-cores (P-cores).
Performance Improvements with Intel Xeon 6 Processors
For the first time, Intel submitted results using Xeon 6 processors with P-cores, achieving about a 1.9x geomean performance improvement compared to the 5th Gen Xeon processors. This performance boost underscores the importance of continuous investment and resource allocation in enhancing AI performance. Over the past four years, Intel has raised AI performance on its Xeon processors by up to 17 times based on MLPerf results.
Benchmarking and AI Inference Capabilities
Intel’s Xeon processors, including the 5th Gen and the new Xeon 6 with P-cores, were tested across six MLPerf benchmarks: ResNet50, RetinaNet, 3D U-Net, BERT, DLRM v2, and GPT-J. These benchmarks showcased the average 1.9x better AI inference performance of the Xeon 6 processors compared to their predecessors. Notably, Intel remains the only server processor vendor submitting CPU results to MLPerf.
Intel’s Commitment to AI Innovation
Pallavi Mahajan, Intel’s Corporate Vice President and General Manager of Data Center and AI Software, highlighted the significant gains in AI performance achieved by Intel over the years. Innovations such as Intel Advanced Matrix Extensions (AMX) play a crucial role in these advancements. The latest MLCommons benchmarks demonstrated that Xeon processors are strong AI server solutions for original equipment manufacturers (OEMs).
OEM Partnerships and Future Prospects
Intel supported five OEM partners—Cisco, Dell Technologies, HPE, Quanta, and Supermicro—in their MLPerf submissions. Each partner showcased the robust AI capabilities of their systems using the 5th Gen Xeon processors. As the demand for AI compute grows, Intel continues to innovate and invest in its CPU roadmap, ensuring that Xeon processors remain at the forefront of AI performance.
Intel plans to provide more detailed information about the Xeon 6 processors with P-cores during a launch event scheduled for September. This ongoing commitment to AI innovation and performance underscores Intel’s leadership in the AI and CPU spaces.