Case Study: NEC accelerates AI research with Supermicro GPU supercomputers

NEC Advances AI Research With Advanced GPU Systems From Supermicro

NEC, a global leader in AI technology and IT systems, required an AI supercomputer with extreme computational power (580 petaflops) to accelerate its research and development, as training complex models like GPT-3 could otherwise take thousands of hours. They sought a trusted partner to deliver a high-performance, customizable, and extensible GPU system to overcome these computational bottlenecks and maintain their competitive edge.

Supermicro provided a solution using their GPU servers, specifically the SYS-420GP-TNAR, which are equipped with NVIDIA A100 80GB GPUs. This advanced infrastructure, featuring high-speed NVLink and NVSwitch technology, enabled massively distributed deep learning. The results were transformative, with NEC reporting that AI development time was reduced by up to 4,600 times compared to a single Tesla V100 GPU, dramatically accelerating their AI research capabilities.

NEC

Takatoshi Kitano

Senior AI Platform Architect