FS has announced its new 1.6T OSFP IHS (Closed Finned Top) optical transceivers, including the OSFP-DR8-1.6T and OSFP-2FR4-1.6T models, specifically targeting high-density artificial intelligence (AI) and high-performance computing (HPC) data center networks using InfiniBand XDR. The company states that these modules are designed to deliver full 1.6T throughput in next-generation NVIDIA GB300 and B300 architectures, addressing both bandwidth and efficiency requirements.
The FS 1.6T OSFP IHS series supports higher bandwidth density via 8x 200G pulse amplitude modulation 4 (PAM4), significantly outpacing mainstream 800G transceivers. This enables larger AI and HPC clusters to be deployed in the same rack space, accommodating the growing scale and connectivity demands of modern compute platforms.
The modules use a Broadcom 3 nm digital signal processor (DSP) for power consumption as low as 26 W. The OSFP-DR8-1.6T module integrates silicon photonics to further decrease heat output and cooling requirements. An integrated heatsink with cooling fins allows these transceivers to operate in both air- and liquid-cooled data center environments.
FS reports that all 1.6T modules are fully tested on NVIDIA Quantum-X800 InfiniBand switches and are verified for low bit error rates, high signal integrity, and long-term durability. The OSFP-DR8-1.6T supports data center distances up to 500 meters for rack-to-system links, while the OSFP-2FR4-1.6T reaches up to 2 kilometers using coarse wavelength-division multiplexing (CWDM4), enabling long-distance inter-switch links required by large AI clusters and advanced machine learning workloads.
“The evolution from 800G to 1.6T is not only about higher speed but also about reshaping the optical interconnect architecture for AI,” said Kyrie Zhang, Senior Product Manager at FS. “FS is committed to delivering proven and ready-to-deploy 1.6T solutions, and our 1.6T OSFP IHS transceiver series gives global professional organizations the reliability and efficiency needed to overcome performance bottlenecks and scale massive clusters.”
FS states that the transceivers are InfiniBand XDR-compliant and interoperable with existing NVIDIA Quantum-X800 networks, aiming to support next-generation rack-scale and row-scale AI and HPC data center deployments.
Source: FS
