NVIDIA MMA4Z00‑NS 2×400G Twin-Port 800Gb/s OSFP SR8 Multimode Transceiver for HPC and Data Centre Networks

NVIDIA MMA4Z00‑NS is a high-performance 800 Gb/s twin-port OSFP transceiver designed for modern data center and HPC environments. Each port delivers 400 Gb/s over multimode fiber with SR8 optics, providing low-latency, high-bandwidth connectivity for demanding workloads.

Compatible with Quantum-2 switches, DGX-H100 systems, ConnectX-7, and BlueField-3 DPUs, this module supports both Ethernet and InfiniBand protocols. Its dual thermal variants, low-power mode, and hot-pluggable design ensure flexible deployment and efficient operation in dense, high-speed networks.

KEY FEATURES:

• Twin-Port 800G OSFP Module: 2 × 400G lanes for high-bandwidth spine-leaf, AI, and HPC networks.
• 4 × 100G PAM4 Lanes per Port: Ultra-low latency and stable 400G throughput.
• Reach: 50 m over OM4 / 30 m over OM3 for short-reach multimode fibre in data centres.
• Connector: MPO-12/APC for reliable, low-loss high-density fibre connections.
• Thermal Variants: Finned-top (air-cooled) or Flat-top (liquid-cooled) compatible.
• Power Efficiency: Maximum 17 W; low-power sleep mode (~2 W).
• Supply Voltage: 3.3V single-supply operation compliant with OSFP standards.
• Certification: NVIDIA Spectrum certified for full compatibility with 400G switches.
• Signal Integrity: DSP and 850 nm VCSEL optics maintain consistent performance.
• Hot-Pluggable & Safety: Class 1 laser, RoHS-compliant, easy to install or replace.
• Applications: Ideal for AI/ML, HPC, and cloud clusters with reliable high-bandwidth links.
• Handling Guidance: Dust-cap protection and fibre cleaning recommended for optimal performance.

TECHNICAL SPECIFICATIONS:

APPLICATIONS:

• Data Centre Interconnects: High-speed spine-to-spine and leaf-to-leaf optical links for large-scale networks.
  

• GPU & AI/ML Clusters: Ultra-low latency, high-bandwidth connectivity for DGX, Quantum-2, and other GPU-accelerated systems.
  

• High Performance Computing (HPC): Optimised for InfiniBand NDR fabrics and workloads requiring reliable 400G links.
  

• Cloud & Hyperscale Networks: Supports scalable 400G infrastructure for AI, big data, and enterprise cloud deployments.
  

• Switch-to-Switch / Spine-Leaf Fabrics: Predictable throughput and minimal congestion in dense networking environments.
  

• Short-Reach Multimode Fibre Applications: Designed for OM3/OM4 links up to 30–50 m, ideal for top-of-rack to spine connections.

ADVANCED FEATURES:

Dual‑Engine Twin‑Port Design:

The MMA4Z00-NS integrates two fully independent 400 Gb/s transceiver engines in a single OSFP form factor, enabling a combined 800 Gb/s capacity. This twin‑port architecture doubles the effective bandwidth density per cage and lowers port count in high-density switch designs. Moreover, each port operates via separate MPO‑12/APC connectors, giving flexibility in cabling and network layout. 

SR8 Parallel Multimode Optics & Modulation:

Using SR8 optics (which equates to 2 × SR4), the module carries 8 parallel channels employing 100 G‑PAM4 modulation to reliably transport 400 Gb/s per port. This design helps maintain signal integrity and low bit error rates over multimode OM3 / OM4 fibre while optimising for high spectral efficiency.

Protocol Flexibility with Auto‑Detection:

The transceiver supports both Ethernet (400 GbE) and InfiniBand NDR, and it auto-configures the protocol on link-up, making deployment across different NVIDIA networking platforms seamless. This automatic adaptation removes the need for manual protocol selection or firmware juggling, simplifying operations in mixed‑protocol environments.

Flexible Fiber Splitting & Breakout:

The module supports straight 400 Gb/s connections, or 200 Gb/s via 1:2 fibre splitters by disabling two optical lanes and switching to lower power mode. The result is reduced power (≈ 5.5 W) when using split‑end cables. This flexibility allows network architects to save cost or power in dense deployments by using fewer lanes per link.

Thermal Variants for Different Cooling Environments: 

There are two mechanical/thermal versions: finned‑top OSFP for air‑cooled switches and flat‑top OSFP for liquid‑cooled systems (e.g., DGX H‑100). The finned-top design improves heat dissipation in high-density switch cages, while the flat-top version fits compact or liquid-cooled form factors.

Low-Power Sleep Mode for Energy Saving:

When not in active use, the transceiver can enter a “sleep” state, consuming as little as 2.0 W, significantly reducing heat and power draw. At full activity, it draws up to 17 W, making it efficient for continuous 400G / 800G operation without overheating.

Diagnostic & Monitoring Capabilities:

Supports CMIS 4.0 (Common Management Interface Specification), which enables detailed real-time monitoring of critical parameters like optical power, temperature, voltage, and bias current. This allows predictive maintenance and helps avoid link failures by tracking module health metrics over time.

Safety & Compliance:

• Certified to Class‑1 laser safety, making it safe for hot-plug operations in data centre environments.

• Fully RoHS-compliant, ensuring the module adheres to environmental regulatory standards. 

• Compliant with OSFP MSA (Multi-Source Agreement) and supports CMIS 4.0 for standardised management.

Optimised for NVIDIA Platforms:

Engineered specifically for Quantum-2 InfiniBand and Spectrum-4 Ethernet switches, enabling seamless integration and maximum throughput in NVIDIA network fabrics. Also supports connections to NVIDIA DPUs / adapters like ConnectX-7 or BlueField‑3, enabling high-performance switching and compute designs.

Advanced Connectivity Options:

Because of the dual-port design, it supports many innovative configurations  for example, one OSFP module can link two separate devices or provide redundant paths, improving both flexibility and fault tolerance in high-performance data center designs.

ORDERING INFORMATION:

WARRANTY & SUPPORT:

Manufacturers are responsible for warranties, which are based on the purchase date and validity. We make sure the products we sell are delivered on time and in their original condition at Grabnpay. Whether it's physical damage or operational issues, our team will help you find a solution. We help you install, configure, and manage your devices. We'll help you file a warranty claim and guide you every step of the way.

All listed prices reflect standard market rates. Final commercial terms, including volume-based discounts, will be provided based on purchase volume, MOQ, and other applicable criteria. For more information about switches & other accessories, contact our technical support team.