The SFPPlaza NVIDIA/Mellanox Compatible 400G 2xSR4 OSFP Optical Transceiver Module is designed for high-performance InfiniBand HDR and 400Gb/s data center interconnect applications. Operating at an 850nm wavelength, this module integrates eight 50G PAM4 optical lanes to deliver an aggregate data rate of 400Gb/s through two independent 200G SR4 optical interfaces.
Featuring an OSFP IHS/Finned Top form factor, the module is designed primarily for switch-side applications in compatible NVIDIA Quantum-2 InfiniBand platforms, including QM9700 and QM9790 switches. Its integrated finned-top housing supports the thermal requirements of high-density, high-speed switching environments.
The dual-port 2xSR4 architecture provides two separate MPO-12/APC optical interfaces. Each interface carries four 50G PAM4 lanes and supports up to 200Gb/s transmission. The module supports optical links of up to 50 meters over OM4 multimode fiber, making it suitable for short-reach connections within data centers, AI clusters, and high-performance computing environments.
For network deployment, the OSFP-400G-2xSR4 module can connect a compatible NVIDIA InfiniBand switch to two 200G QSFP56 HDR network adapters through two independent MPO-12/APC multimode fiber links. This architecture enables efficient 400G aggregate connectivity while maintaining two separate 200G optical channels. Correct fiber polarity, lane mapping, and switch port configuration are essential for successful link establishment.
Unlike conventional single-port 400G SR4 transceivers, which use four 100G PAM4 lanes through one MPO optical interface, the 400G 2xSR4 architecture uses eight 50G PAM4 lanes distributed across two independent MPO-12/APC interfaces. These different lane rates and optical architectures mean that standard 400G SR4 modules cannot directly interoperate with 400G 2xSR4 modules.
With Digital Monitoring (DOM) support and compliance with applicable OSFP MSA and CMIS specifications, this transceiver helps network operators monitor module operating conditions and maintain reliable optical connectivity. It is a practical solution for NVIDIA InfiniBand HDR switching environments, AI infrastructure, HPC clusters, and high-bandwidth data center networks.
