H3C Again Tops the MLPerf International Authoritative AI Benchmark Test with 25 World No.1 Titles!

The latest AI Inference V3.1 results, which were developed by the international authoritative AI benchmarking organization MLCommons, have recently been released by MLPerf™ benchmarks. Twenty-five semiconductor, server, and algorithm vendors worldwide participated in the evaluation. H3C stood out with its AI servers, securing 25 world No.1 titles, reflecting its strong technical innovation and R&D capabilities in the AI field.

MLPerf™, initiated by Turing Award laureate David Patterson and top academic institutions, holds the highest reputation and participation rate in AI benchmark tests worldwide. It establishes classic model tracks like natural language processing, medical image processing, intelligent recommendations, and provides a fair assessment of the training and inference performance of vendors' hardware, software, and services. The test results serve as an authoritative and effective data guide for users to measure device performance and a touchstone for vendors' technical strength in the AI field. With years of dedication and profound strength, H3C has won a total of 157 championships at MLPerf.

H3C UniServer R5300 G6: A Flagship with Diverse Computing Power in the Intelligent Era

In the AI inference benchmark test, the R5300 G6 server turned in an excellent performance, winning 23 configuration-first titles in both data center and edge scenarios, plus one absolute first. It demonstrated its strong support for large-scale, diversified, and complex computing power scenarios.

In the ResNet50 model track, the R5300 G6 can classify 282,029 images per second, offering efficient and accurate image processing and recognition capabilities.

In the RetinaNet model track, the R5300 G6 can identify targets in 5268.21 images per second, providing computing power support for scenarios like automatic drive, smart retail, and smart manufacturing.

In the 3D-UNet model track, the R5300 G6 can process 26.91 3D medical images per second at a 99.9% accuracy requirement, assisting doctors with fast diagnosis and improving treatment efficiency and quality.

As a flagship for diverse computing power in the intelligent era, the R5300 G6 boasts outstanding performance, flexible architecture, strong expandability, and high reliability. It supports various types of AI accelerator cards, and its CPU and GPU mounting ratios support 1:4 and 1:8. Its five GPU topologies flexibly adapt to the needs of different AI scenarios. Additionally, the R5300 G6 adopts a computing-storage integrated design, supporting up to 10 dual-width GPUs and 400TB of massive storage to meet AI data's storage space requirements.

H3C UniServer R5350 G6: A Hybrid Computing Engine for the Intelligent Era

Meanwhile, with its leading AI system design and full-stack optimization capabilities, the R5350 G6 server secured the configuration-first title in the ResNet50 (image classification) evaluation task in this benchmark test. Compared to its predecessor, the R5350 G6 achieves a 90% performance improvement and a 50% increase in core count. With 12-channel memory, the memory capacity can reach 6TB. Simultaneously, the R5350 G6 supports up to 24 2.5/3.5-inch hard drives, 12 PCIe5.0 slots, and a 400GE network card. It meets AI's demands for massive data storage and high-speed network bandwidth and can be applied to various application scenarios such as deep learning training, deep learning inference, and data analysis.

Each performance breakthrough and record reset demonstrates H3C's commitment to cultivating customer application scenarios and accumulating practical experience and technical capabilities. In the future, H3C will adhere to the strategy of “Dedication, For A Smarter Future”, closely imbed product innovation in AI application scenarios, bringing continuously-evolving intelligent computing power to all industries.