AI & UE

Ultra Ethernet is designed to address the scalability, congestion, and latency limitations of traditional Ethernet and RoCE is supporting large‑scale AI and HPC clusters with up to millions of endpoints.  Enabling reliable, lossless performance on this level requires next-level testing.

The AI & UE Solution Track is for use with Z800o Freya and Z1608 Edun traffic generators. It enables verification, optimization, and robustness testing of UE Link Layer functionality under realistic and demanding load conditions.

This Solution Track is primarily intended for:

• Switch and NIC/GPU developers implementing UEC features
• AI‑intensive data‑center operators planning large‑scale UE deployments
• Vendors targeting HPC/AI clusters that require deterministic, loss‑free transport.

In addition to the standard Ethernet testing offered by XenaManager3 (XM3), the AI & UE Solution Track adds UE‑specific Link Layer behaviors required for validating UEC‑enabled switches and xPUs.

These include:

• Link Layer Retry (LLR)
• Credit‑Based Flow Control (CBFC)
• Link Layer Negotiation (LLDP for UE extensions)
• Stateful UE protocol control and message exchange
• UE error injection, message inspection, and capture

The functionalities cover stateful protocol communication, error injection, and capture capabilities for comprehensive testing.

With these capabilities, the AI & UE Solution Track transforms Z800o Freya and Z1608 Edun modules into UE‑compliant endpoints, enabling highly realistic and standards‑aligned traffic behavior for early development, interoperability validation, and performance stress testing.

The AI UE Solution Track follows the UEC standard and will in the future be expanded with capabilities for testing e.g. Packet Rate Improvements, Telemetry, Negotiation Enhancements, and Idle Reduction.