KR6288-E3 is an advanced AI system featuring 2 AMD EPYC™ 9005 processors and supporting NVIDIA HGX H200 8-GPU. This server delivers industry-leading 32 PFlops of AI performance and lightning-fast CPU-to-GPU interconnect bandwidth, with the H200 Transformer Engine supercharging training speeds for GPT large language models.

Its optimized power efficiency and modular design with flexible configuration makes it ideal for AI model training and inference, AI start-ups, AI cloud services, and converged scenarios that take advantage of the AMD platform’s advanced core count and high memory bandwidth.

Unprecedented AI Performance

• Powered by NVIDIA HGX H200 8-GPUs in 6U, TDP up to 700W
• 2x AMD EPYC™ 9005 processors
• 32 PFlops of industry-leading AI performance
• H200 Transformer Engine delivers supercharged training speed for GPT large language models

Leading Architecture Design

• Lightning-fast CPU-to-GPU interconnect bandwidth
• Ultra-high scalable inter-node networking with up to 4.0 Tbps non-blocking bandwidth 
• Optimized cluster-level architecture with 8:8:2 ratio of GPU to compute network to storage network

Optimized Energy Efficiency    

• Low air-cooled heat dissipation overhead and high power efficiency
• 54V, 12V separated power supply with N+N redundancy reducing power conversion loss

Flexible Configurations for AI Scenarios

• Fully modular design and flexible configurations satisfy both on-premises and cloud deployment
• Easily harness large-scale model training, such as GPT-3, MT-NLG, stable diffusion and Alphafold
• Diversified SuperPod solutions accelerating the most cutting-edge innovation including AIGC, AI4Science and Metaverse