Should an AI cluster use 400G or 800G?
Choose by topology, host generation and bandwidth target. Both rates may coexist across node, Leaf and Spine layers.
AI Compute and GPU Clusters
Plan AI cluster optics by GPU, Leaf and Spine position, then match form factor, reach, fiber, thermal limits, FEC and breakout mode.
Selection FAQ
Choose by topology, host generation and bandwidth target. Both rates may coexist across node, Leaf and Spine layers.
Evaluate VR8/SR8 for MMF, DR8 for 500m-class parallel SMF and 2xFR4 for 2km-class SMF, subject to the exact host and PMD.
Validate host coding, FEC, errors, optical power, power draw, temperature, link flap, fiber mapping and long-duration stability.
Typical Scenarios
Match the NIC, switch port, rack distance and cable path.
Plan 400G and 800G links by reach, fiber and port density.
Validate host breakout, lane mapping, cabling, FEC and every remote endpoint.
Selection Guide
Recommended Products
Related Solutions
100G–800G modules and links
View solution →Servers, racks and breakout
View solution →Multi-vendor samples and rollout
View solution →Selection Guides
Compare 50m, 100m, 500m and 2km PMD directions for AI fabrics.
Read guide →Record loss, margin, receiver sensitivity and overload on the final path.
Read guide →Prepare host, FEC, sample and batch-acceptance fields before RFQ.
Read guide →