NVIDIA/Mellanox MMS4A50-XM (980-9IAS0-00XM00) Compatible 1.6T 2xFR4/FR8 OSFP224 IHS/Closed Finned Top PAM4 Broadcom 3nm DSP (Sian3 | BCM83628) 1310nm 2km SMF DOM Dual LC Duplex InfiniBand XDR Silicon Photonics Transceiver Module for Quantum-X800 Air-Cooled Switches

Name: NVIDIA/Mellanox MMS4A50-XM (980-9IAS0-00XM00) Compatible 1.6T 2xFR4/FR8 OSFP224 IHS/Closed Finned Top PAM4 Broadcom 3nm DSP (Sian3 | BCM83628) 1310nm 2km SMF DOM Dual LC Duplex InfiniBand XDR Silicon Photonics Transceiver Module for Quantum-X800 Air-Cooled Switches
Brand: Naddod
SKU: 102790
Price: 2299.00 USD
Availability: OnlineOnly
Rating: 5.0 (18 reviews)

Question 1

Q:

What are the practical differences of 3nm DSP over common 5nm DSP solutions in real use?

Answer

A:

NADDOD's 1.6T transceiver features Broadcom's 3nm DSP (BCM83628). It boasts lower power consumption and better overall performance than 5nm DSP solutions, while significantly lowering the port failure rate and further enhancing link stability in actual deployment. Tested on the NVIDIA Quantum-X800 Q3400 switch, the 3nm solution delivers higher energy efficiency and stability, ensuring efficient and reliable operation of large AI clusters.

Question 2

Q:

When building or upgrading an AI cluster, what application scenarios and computing scale require this 1.6T XDR transceiver?

Answer

A:

The 1.6T XDR transceiver is designed for AI computing scenarios with high network performance requirements, and is suitable for medium to ultra-large scale AI clusters. For medium and large-scale model training and inference scenarios, it meets the requirements of efficient data exchange and computing power expansion. For ultra-large GPU clusters and factory-level AI deployments, it delivers higher bandwidth density, ensuring smooth cluster operation and supporting stable long-term expansion.

Question 3

Q:

I plan to purchase about 1000 XDR transceivers in one batch for cluster upgrade, what is the current production capacity and expected delivery time?

Answer

A:

NADDOD's XDR transceivers are not sample or R&D stage products, and have been deployed in batches in real AI clusters with a stable mass production and delivery system. We have sufficient supply capacity with short delivery time, and can fully support large-scale project-level delivery. For your specific procurement volume and deployment plan, you may directly contact your dedicated account manager or learn about the detailed delivery time via sales@naddod.com.

Question 4

Q:

What's the difference between EML and SiPh applied in 1.6T XDR transceivers?

Answer

A:

EML features high stability and mature manufacturing technology, and is suitable for long-distance transmission scenarios. SiPh has prominent advantages of ultra-low power consumption, ultra-low latency and high-density adaptability, making it an ideal choice for high-density AI cluster deployment. These two technologies differ in chip structure, modulation mode and applicable scenarios. For more detailed comparisons, please refer to our technical blog on (EML vs. SiPh).

Question 5

Q:

What's the difference between LC‑interface 1.6T optical modules and MPO‑interface modules?

Answer

A:

MPO connectors are used in multimode or parallel single‑mode setups such as 2×SR4/SR8 or 2×DR4/DR8, which support multiple optical lanes and are well‑suited to short‑reach, high‑speed interconnects. LC connectors, on the other hand, are used for 2×FR4 (CWDM4) – a wavelength‑multiplexed point‑to‑point architecture optimized for medium‑to‑long‑reach links. 2×FR4 uses dual‑duplex LC interfaces and is geared toward point‑to‑point interconnects up to several thousand kilometers.

Question 6

Q:

Can the IB 1.6T 2×FR4 OSFP224 module be used with the NVIDIA QM9700 switch?

Answer

A:

No. The QM9700 switch requires 8×100 G SerDes (aligned with 800G DR4 modules), whereas the 1.6 T 2×FR4 OSFP224 module over FR4 uses 8×212 G SerDes—this discrepancy in SerDes rates makes the two incompatible.

Question 7

Q:

What are the typical application scenarios for the IB 1.6T 2×FR4 OSFP224 module?

Answer

A:

These modules are primarily used to interconnect NVIDIA Quantum‑X800 Q3400‑RA switches in high‑performance InfiniBand clusters.

Question 8

Q:

Is it possible to use only one 800G channel and disable the other?

Answer

A:

Yes, but only via host‑side CMIS configuration or driver/firmware control that allows per‑lane activation or dormancy. With host software intervention you can activate only one FR4 sub‑link (800G) while the other remains idle. The module hardware must support separate CMIS control bits to gate channels individually.

Question 9

Q:

What upgrades and changes have been made from 800G 2×FR4 optical modules to 1.6 T 2×FR4 OSFP224 optical modules?

Answer

A:

It supports PAM‑4 modulation at up to 200 Gbps per lane (an upgrade from the previous ~100 G PAM‑4 lane‑rate). This doubles the per‑link bandwidth, increases port density, and often improves power efficiency and thermal performance.