“Today, Clockwork builds a software layer that focuses on optimizing the communication between GPUs in large clusters that are then used for AI workloads,” Vasudevan told me. “As you well know, AI workloads are among the most distributed and most demanding distributed applications in history. A lot of how well the workload performs depends on how effective the communication is between GPUs. What Clockwork focuses on is a set of software building blocks that allow you to get three things that ultimately lead to higher AI efficiency.”
These…
