Overview
Explore
Expertise
What's new
Compare the Cloud: WWT and DDN publish a storage design for 50MW AI factories that targets 2.5GB/s of reads per GPU
WWT and DDN unveil a storage reference design built to support next-generation AI factories.
WWT and DDN design storage to help AI factories fit more GPUs within their power limits
Reference design for 50 MW campuses on NVIDIA Vera Rubin targets more than 2.5 GB/s of read throughput per GPU, and lets inference nodes share KV cache, the working memory built for each prompt