 
|
MMA4Z00-NS Mellanox Transceiver 800Gb/S Twin Port OSFP 2x400Gb/S Multimode SR8 50m
Dettagli:
| Marca: | Mellanox |
| Numero di modello: | Ricetrasmettitore ottico MMA4Z00-NS di Mellanox |
| Documento: | MMA4Z00-NS.pdf |
Termini di pagamento e spedizione:
| Quantità di ordine minimo: | 1 pz |
|---|---|
| Prezzo: | Negotiate |
| Imballaggi particolari: | Scatola esterna |
| Tempi di consegna: | Basato sull'inventario |
| Termini di pagamento: | T/T |
| Capacità di alimentazione: | Fornitura per progetto/batch |
|
Informazioni dettagliate |
|||
| Condizione: | Nuovo e originale | Disponibilità: | Azione |
|---|---|---|---|
| Tecnologia: | Infiniband ed Ethernet | Velocità massima: | 800 GB/s |
| Lunghezza: | 50m | Intervallo di temperatura: | Da 0°C a +70°C |
Descrizione di prodotto
2x400Gb/s Multimode SR8 | Up to 50m Reach | Finned-Top for Quantum-2 & Spectrum-4
High-density 800G solution: NVIDIA’s MMA4Z00-NS is an InfiniBand and Ethernet 800Gb/s Twin‑port OSFP transceiver integrating two independent 400G SR8 engines in a single OSFP form factor. Using dual MPO‑12/APC connectors and 100G‑PAM4 modulation over 8 multimode fibers, it delivers 2x400G optical links up to 50 meters (OM4). Ideal for Quantum‑2 and Spectrum‑4 switches, enabling 64 ports of 400G in just 32 cages.
The NVIDIA MMA4Z00-NS (800Gb/s Twin‑port OSFP SR8) is a breakthrough transceiver that integrates two parallel 400G multimode optical engines into a single OSFP plug. Each engine uses 4 lanes of 100G‑PAM4 (SR8) and connects via independent MPO‑12/APC connectors. This unique twin‑port design allows network architects to maximize switch faceplate density: a single OSFP cage in Quantum‑2 or Spectrum‑4 switches delivers two 400G optical interfaces. With built‑in InfiniBand and Ethernet firmware auto‑detection, the transceiver works seamlessly across NDR InfiniBand and 400GbE environments. Designed for short‑reach multimode links up to 50 meters, it is the optimal choice for spine‑leaf topologies, GPU clusters, and high‑performance data center interconnects.
Two internal transceiver engines in one module: 2x400G using dual MPO-12/APC connectors; reduces switch port count while doubling density.
8 channels of 100G‑PAM4 electrical and optical modulation per transceiver; aggregate 800Gb/s.
850nm VCSEL array; reach up to 30m (OM3) / 50m (OM4) with two patch panels assumed.
Supports straight cables (2x400G) or 1:2 splitter cables for 2x200G to four adapters/DPUs.
17W max, 2.0W low‑power sleep mode; CMIS 4.0 management interface with DDM and advanced diagnostics.
Auto‑detects InfiniBand or Ethernet based on switch protocol; simplifies inventory.
Built around NVIDIA’s twin‑port architecture, the MMA4Z00‑NS uses two independent 4‑channel optical subassemblies, each driving 100G‑PAM4 signaling over OM3/OM4 multimode fiber. The transceiver integrates 850nm VCSELs, PIN photodiodes, and precision DSP for robust signal integrity. The MPO‑12/APC angled physical contact connectors minimize back reflection, enabling reliable performance in dense optical interconnects. The module supports adaptive Tx equalization, programmable Rx output amplitude, and built‑in PRBS generators for host/line loopback testing. Digital diagnostic monitoring (DDM) provides real‑time temperature, voltage, bias current, Tx/Rx power per lane.
- Switch‑to‑Switch: Connect two Quantum‑2 or Spectrum‑4 switches at 800Gb/s using two straight fibers (2x400G). Alternatively, link a single Twin‑port transceiver to two different switches, forming independent 400G links.
- Switch‑to‑Adapter/DPU: Connect one Twin‑port transceiver to two 400G ConnectX‑7 adapters or BlueField‑3 DPUs (OSFP or QSFP112) using straight cables.
- Switch‑to‑four 200G devices: Deploy 1:2 splitter cables to split each 400G port into 2x200G links, enabling four 200G connections per Twin‑port transceiver.
- DGX H100 Integration: Flat‑top version (MMA4Z00‑NS‑FLT) connects Cedar‑7 mezzanine cards to Quantum‑2/Spectrum‑4 switches, simplifying GPU‑to‑switch cabling and reducing transceiver count vs. previous DGX systems.
The MMA4Z00‑NS is validated for NVIDIA end‑to‑end systems, guaranteeing optimal interoperability. Compatible platforms include:
- NVIDIA Quantum‑2 (NDR) InfiniBand switches – 32 OSFP cages supporting 64x400G ports with finned‑top modules.
- NVIDIA Spectrum‑4 Ethernet switches – 32/64 OSFP cages delivering 64‑128 400G ports.
- NVIDIA ConnectX‑7 adapters (OSFP and QSFP112 form factors) and BlueField‑3 DPUs.
- DGX H100 systems (with flat‑top variant) – eight ConnectX‑7 integrated into Cedar‑7 cards.
- Standard CMIS 4.0 host platforms compliant with OSFP MSA Rev 1.12.
| Parameter | Value / Range |
|---|---|
| Data Rate | 800Gb/s aggregate (2x400Gb/s) |
| Form Factor | OSFP (Octal Small Form Factor Pluggable) Twin‑port |
| Optical Connector | Dual MPO‑12/APC (8‑fiber per port, 12‑fiber ribbon footprint) |
| Fiber Type & Distance | OM3: up to 30m / OM4: up to 50m (with two patch panels assumed) |
| Wavelength | 850nm VCSEL (λc 844–863nm) |
| Modulation | 100G‑PAM4 per lane, 8 lanes total per module |
| Power Supply | 3.3V single supply; Max power 17W, Low‑power sleep 2.0W |
| Operating Case Temp | 0°C to 70°C |
| Management Interface | CMIS 4.0 (I²C), OSFP MSA compliant |
| Laser Safety | Class 1 Laser Product, RoHS compliant |
| Part Numbers | MMA4Z00-NS (finned‑top) / MMA4Z00-NS-FLT (flat‑top) |
| Part Number | Description | Top Shell | Target System |
|---|---|---|---|
| MMA4Z00-NS | 800Gb/s OSFP SR8 twin‑port, 2xMPO, 50m multimode | Finned‑top | Quantum‑2 / Spectrum‑4 air‑cooled switches |
| MMA4Z00-NS-FLT | 800Gb/s OSFP SR8 twin‑port, 2xMPO, 50m multimode | Flat‑top | DGX H100, liquid‑cooled systems |
Recommended Fiber Cables (NVIDIA crossover Type‑B): Straight fibers MFP7E10-Nxxx (3m to 50m) for 2x400G links; 1:2 splitter cables MFP7E20-Nxxx (3m to 50m) for 2x200G per split port. Both cables in a Twin‑port transceiver must be same type (both straight or both splitter) but lengths can differ.
- Ultimate Port Density: Doubles 400G port count in switch fabric – reduces switch count and cabling complexity.
- End‑to‑End NVIDIA Validation: Rigorous production testing ensures out‑of‑the‑box reliability and performance in Quantum‑2/Spectrum‑4 and ConnectX‑7 ecosystems.
- Operational Flexibility: One transceiver SKU supports InfiniBand and Ethernet, automatically enabling correct protocol.
- Power Efficiency: 17W max with automatic power reduction when using splitter cables (400G transceivers drop to 5.5W).
- Advanced Diagnostics: Per‑lane Tx/Rx power, temperature, bias monitoring plus PRBS loopback for link validation.
Starsurge Group provides comprehensive pre‑sales and after‑sales support for NVIDIA networking solutions. Our technical team assists with compatibility validation, deployment planning, and troubleshooting. All transceivers are backed by warranty and come with original quality assurance. Global shipping, multilingual support, and custom configuration services available. Contact our experts for BOM optimization and volume pricing.
- Class 1M laser product – avoid direct exposure to laser light. Use protective gear when handling.
- Keep dust caps on optical ports when not in use. Clean optical connectors before mating; 80% of link issues originate from dirty connectors.
- Observe ESD precautions – damage may occur from electrostatic discharge.
- Do not exceed absolute maximum ratings (storage -40°C to 85°C, supply voltage 3.6V max).
- Use only NVIDIA‑approved crossover Type‑B fiber cables for direct transceiver‑to‑transceiver connection to ensure correct polarity and performance.
Hong Kong Starsurge Group Co., Limited is a technology‑driven provider of network hardware, IT services, and system integration solutions. Founded in 2008, the company serves customers worldwide with products including network switches, NICs, wireless access points, controllers, cables, and related equipment. Backed by an experienced sales and technical team, Starsurge supports industries such as government, healthcare, manufacturing, education, finance, and enterprise. With a customer‑first approach, Starsurge focuses on reliable quality, responsive service, and tailored solutions that help clients build efficient, scalable, and dependable network infrastructure.
As an authorized channel partner, Starsurge supplies genuine NVIDIA transceivers, cables, and switches with full logistic support and technical expertise.
- Confirm switch model (Quantum‑2 / Spectrum‑4) and required top‑shell type (finned or flat).
- Choose OM3/OM4 multimode cable lengths and decide straight or 1:2 splitter configuration.
- Verify that both cables connected to the Twin‑port transceiver are of same type (both straight or both splitter).
- Check power budget: 17W per transceiver, ensure switch cooling capacity.
- For splitter usage, confirm adapter/DPU supports 200G (2‑lane) operation.
- MMA1Z00-NS400 – 400Gb/s QSFP112 SR4 transceiver for ConnectX‑7 / BlueField‑3
- MMA4Z00-NS-FLT – Flat‑top variant for DGX H100 & liquid‑cooled deployments
- MFP7E10-N0xx – Straight Type‑B multimode crossover fiber cables (3m–50m)
- MFP7E20-N0xx – 1:2 splitter crossover cables (2x200G per split)
- Quantum‑2 QM9700 Switch System – 64 ports of 400G NDR InfiniBand