Rambus Bets on Time Division Multiplexing to Fix PCIe 7.0 for AI Workloads, as GPUs Starve for Data

Rambus introduced PCIe 7.0 switch IP for AI and data center infrastructure.

The product uses time division multiplexing to schedule traffic across shared links more efficiently.

It is designed to improve bandwidth utilization, lower latency, and scale AI, cloud, and HPC workloads.

The IP aims to help move data between CPUs, GPUs, accelerators, and NVMe storage more effectively.

Rambus says the technology also supports future pooled and disaggregated compute architectures.