NVIDIA/Mellanox MMA4Z00-NS (980-9I510-00NS00) Compatible 800G 2xSR4 OSFP IHS/Finned Top 8x100G PAM4 Broadcom DSP & Broadcom VCSEL 850nm 50m MMF DOM Dual MPO-12/APC InfiniBand NDR Optical Transceiver Module

OSFP-800G-2xSR4-IB: Frequently Asked Questions

Q: What type of optical fiber and patch cables are required for the 800G Twin-Port NDR 2xSR4 optical transceiver?

A: This 800G Twin-Port 2xSR4 OSFP transceiver uses two independent MPO-12/APC optical connectors and requires multimode fiber (MMF) patch cables. For a 2x400G-to-2x400G connection, two OM4 MPO-12/APC patch cords are required, with one cable for each 400G optical port. For a 2x400G-to-4x200G breakout configuration, two compatible OM4 MPO-to-2xMPO-4/APC breakout patch cables may be required, depending on the remote transceiver architecture and fiber polarity. The corresponding NADDOD cable models are M4MPOA12FB and M4MPOA2x4F, respectively. Always verify connector polarity, fiber mapping, and remote-end port configuration before deployment.

---

Q: What is the modulation method and optical lane architecture of the 800G Twin-Port NDR 2xSR4 transceiver?

A: This module uses an 8-lane 100G PAM4 electrical interface and two independent 4-lane 100G PAM4 optical engines. Each optical port provides 400Gb/s bandwidth using 4x100G PAM4 signaling, delivering 800Gb/s aggregate bandwidth across both ports. This dual-port architecture enables flexible 2x400G connectivity for high-performance InfiniBand NDR and compatible Ethernet networking applications.

---

Q: What are the differences between Flat Top and Finned Top OSFP optical transceivers?

A: Flat Top and Finned Top refer to different thermal management designs of OSFP transceivers. Flat Top modules feature a low-profile upper surface, typically relying on the host system’s dedicated cooling infrastructure and airflow design for thermal management. Finned Top modules incorporate an integrated heat sink with cooling fins to increase the heat dissipation surface area and improve thermal performance. The Finned Top design is particularly suitable for high-density, high-speed networking environments with demanding thermal requirements. The appropriate version should be selected according to the host switch or network adapter’s mechanical clearance and cooling specifications.

---

Q: What are the advantages of using Broadcom DSP and Broadcom VCSEL technology in this 800G optical transceiver?

A: The combination of Broadcom DSP and Broadcom VCSEL technology provides an integrated solution for high-speed multimode optical transmission. The DSP performs advanced signal processing to help maintain signal integrity and reliable link performance, while the VCSEL-based optical engine enables efficient 850nm multimode transmission. This architecture is designed to support stable 800G connectivity with low bit error rates and reliable operation, making it suitable for demanding AI training, HPC, and large-scale data center networking applications. Actual BER and power performance depend on the module specifications and operating conditions.

---

Q: My network card is a ConnectX-6 QSFP56 200G adapter. Can I use an 800G Twin-Port NDR 2xSR4 OSFP module to connect directly to a QSFP56 200G SR4 transceiver?

A: Not directly in a conventional 4x200G breakout configuration. This 800G Twin-Port 2xSR4 module provides two independent 400G optical ports, each using 4x100G PAM4 signaling. A QSFP56 200G SR4 transceiver uses 4x50G PAM4 signaling, so the electrical and optical lane rates must be properly supported and mapped by the host interface and breakout configuration. For a 200G connection, we recommend using an appropriate 200G-capable transceiver and supported port configuration. Alternatively, a compatible ConnectX-7 QSFP112 400G interface with a suitable 400G SR4 transceiver may be used in a supported 400G-to-200G breakout configuration. Always verify the adapter’s port breakout capabilities, firmware, and supported optical specifications before deployment.

---

Q: Can this 800G 2xSR4 optical transceiver be used as an 800G SR8 module?

A: Although both modules provide 800Gb/s aggregate bandwidth and use eight 100G PAM4 optical lanes, their optical architectures are different. The 800G 2xSR4 module contains two independent 400G SR4 optical interfaces, each with its own MPO-12/APC connector. An 800G SR8 module uses a single 800G optical interface with eight optical lanes through one multi-fiber connector. Therefore, the two modules have different port configurations and optical connectivity and should not be treated as directly interchangeable.

---

Q: What is the maximum transmission distance of the NVIDIA MMA4Z00-NS compatible 800G 2xSR4 OSFP module?

A: This module operates at a nominal wavelength of 850nm and supports transmission distances of up to 30m over OM3 multimode fiber and 50m over OM4 multimode fiber. It uses two MPO-12/APC optical connectors, with each connector serving an independent 400G SR4 optical port. OM4 fiber is recommended when the full 50m transmission distance is required.

---

Q: How many 400G ports can be supported by NVIDIA Quantum-2 switches using 800G Twin-Port 2xSR4 transceivers?

A: The Twin-Port 2xSR4 architecture provides two independent 400G optical interfaces per OSFP module. In a compatible 32-cage Quantum-2 switch configuration, 32 modules can provide up to 64 optical 400G ports, subject to the switch’s supported port configuration and system architecture. This high-density design helps optimize front-panel port utilization and fiber connectivity for large-scale InfiniBand NDR networks.