Everyone expects GPUs like Nvidia's Tesla P100 in purpose-built AI servers, but optimizing memory access is also a major issue. IBM is using Nvidia's NVLink to help with that bottleneck.
IBM’s new AI-friendly server adopts Nvidia’s NVLink for faster memory
Everyone expects GPUs like Nvidia's Tesla P100 in purpose-built AI servers, but optimizing memory access is also a major issue. IBM is using Nvidia's NVLink to help with that bottleneck.
