Product Overview: 

Designed for the most demanding AI/ML training clusters, our 800G OSFP LPO removes the power-hungry DSP (Digital Signal Processor), relying instead on high-linearity analog components. The result? A 50% reduction in power consumption and near-zero latency—critical for synchronizing thousands of GPUs.

Technical Specifications:

• Architecture: DSP-free Linear Drive (Direct Drive by Host ASIC)

• Power: Industry-leading <7W per module

• Latency: Reduced by ~100ns compared to standard retimed transceivers.

• Application: Intra-rack and inter-rack GPU fabric (up to 500m).

Market Insight:

• Green Data Center: Significant TCO savings in cooling and electricity for hyperscale deployments.

• Compatibility Note: Requires LPO-enabled host switches/NICs (e.g., Arista, NVIDIA Blackwell).