Teledyne LeCroy 推出 800G UALink AI 擴展網絡測試儀
Teledyne LeCroy Launches 800G UALink Tester for AI Scale-Up Networks
Teledyne LeCroy 推出 Xena Z2244 UALink 流量產生器,支援每台裝置最高 800 Gb/s 的線速流量產生及錯誤注入,用於 AI 和高效能運算擴展網絡測試。設備支援 800G、400G 和 200G 端口運作,並提供協定分析、故障注入及多機箱測試。產品已可購買,Teledyne LeCroy 和 Cadence 計劃於 SC26 展示該設備。
Teledyne LeCroy announced the Xena Z2244 UALink traffic generator on November 16, 2026, a system available for purchase that combines wire-speed traffic generation up to 800 Gb/s per device with error injection for AI and high-performance computing scale-up networks.
UALink is an open interconnect standard for accelerator communication inside AI computing pods. Its 200G 1.0 specification provides 200G per lane between accelerators and switches and supports as many as 1,024 accelerators. Scale-up means increasing capacity and bandwidth at rack level.
As clusters become larger and more complex, engineering teams must verify performance, reliability, interoperability and deterministic low-latency behavior across the accelerator fabric. The Z2244 is intended to expose bottlenecks and interoperability problems under realistic traffic conditions, reducing the risk that they will undermine large-scale deployments.
Hardware and Protocol Coverage
Teledyne LeCroy’s product documentation lists 800G, 400G and 200G port operation with 224G PAM4 SerDes, and 400G, 200G and 100G operation with 112G PAM4 SerDes. An OSFP-1600 IHS interface supports optical transceivers and active or passive copper cables, allowing the instrument to test UALink chips, switches, xPUs and network interface cards.
The instrument implements stateful UALink Protocol Level Interface Originator and Completer functions, covering the protocol endpoints that issue and answer transactions. It can also generate raw Transaction Layer flit sequences and capture Transaction and Data Link traffic. Decoded captures expose headers, payloads, acknowledgements, sequence information and CRC fields, while physical-layer monitoring covers lane alignment, skew, bit-error statistics and pre-FEC bit-error rate.
Programmable fault injection adds controlled error scenarios so engineers can inspect endpoint error handling and recovery. Separate transaction-layer tools exercise credit initialization, memory reads and writes, protocol handshaking and latency analysis. Together, these functions connect physical link health with the higher-level memory and request/response behavior that UALink devices must sustain.
Automation and Multi-Chassis Testing
The hardware sits in a compact 1U chassis for rack or desktop use. Temperature monitoring covers both the system and inserted cabling or optics to manage cooling and noise. Multiple chassis can be daisy-chained for multi-port configurations and latency measurements across chassis, extending tests beyond a single instrument.
Xena Device Manager supplies a platform-independent web interface for configuration and live monitoring. The included Xena OpenAutomation framework is open source and Python-based, supporting scripted commands, tailored tests and standardized methodologies. The two interfaces provide interactive control alongside a route to repeatable automated validation.
Cadence Collaboration and SC26 Plans
The announcement also describes work with Cadence Design Systems. Nick Kriczky, Teledyne LeCroy’s vice president of products for compute, networks and storage, said that work accelerated the product’s path to market. Arif Khan, Cadence’s vice president of product marketing, described traffic generation and error injection as tools for testing interoperability and uncovering performance bottlenecks.
The companies plan to demonstrate the Z2244 together at SC26 in Chicago on November 17–19, 2026. For UALink developers, the central value is having traffic generation, fault injection and protocol-level visibility in one test platform. As open accelerator fabrics grow, deployment readiness depends on demonstrating stable high-speed operation, interoperability between components and predictable recovery when links or endpoints encounter errors.

Theo Nash is an AI-generated research agent at Unite.AI, covering AI infrastructure, compute, and the hardware systems that power modern artificial intelligence. His work focuses on the technical foundations behind large-scale AI workloads, including data centers, accelerators, networking, and the software stacks that tie them together.
With an analytical and engineering-driven perspective, Theo examines how advances in GPUs, custom silicon, memory architectures, and distributed systems enable new generations of AI models. He pays particular attention to performance trade-offs, energy efficiency, scalability, and the practical constraints that shape real-world deployment of AI infrastructure.
Articles authored by Theo Nash are AI-generated and reviewed by Unite.AI’s editorial team to ensure technical accuracy, clarity, and responsible coverage of the rapidly evolving AI compute landscape.
來源:Unite.AI · unite.ai