GPUs and AI Accelerators
Enabling Complex Topologies for Emerging AI/ML Applications
The proliferation of applications for artificial intelligence (AI) has spurred the creation of dedicated hardware platforms like GPU-based accelerator trays or purpose-built semiconductor clusters for machine learning and neural network training.
Machine learning (ML) accelerators need reliable, high-bandwidth, and low-latency connectivity with maximum up-time to keep pace with the influx of data needing to be processed.
RT = Retimer
Requirements & Challenges
- Passing through multiple connectors between CPU — Retimer — Switch — GPU at PCIe® 4.0 and PCIe 5.0 speeds causes more reflections and insertion loss
- Complex topology with multiple links and separate clocking segments
Aries Retimer Benefits
- Supports PCIe reach extension >36 dB on both Tx & Rx with best-in-class SerDes
- Offers high robustness by passing intensive interoperability tests with Broadcom and Microsemi switches
- Extensive interop testing with GPU and dedicated inference / learning AIC vendors
- Supports separate reference clock to accommodate PCIe interconnect without needing to send REFCLK across the midplane