MCP7H60-W002 NVIDIA/Mellanox Compatible, 400G QSFP-DD to 2x200G QSFP56 Passive Breakout DAC Cable - 2m
Warranty: Lifetime
NVIDIA/Mellanox Compatible 400G QSFP-DD to 2x200G QSFP56 Passive Breakout Direct Attach Copper (DAC) Cable - 2 Meters
This third-party passive breakout DAC cable is fully compatible with NVIDIA/Mellanox LinkX series (e.g., MCP7H60-W002 equivalents), providing a cost-effective fan-out solution: one 400G QSFP-DD port to two 200G QSFP56 ports. Perfect for hybrid AI training clusters, HPC environments, and hyperscale data centers mixing 400G and 200G infrastructure.
Engineered to MSA standards, it supports full 400Gbps (split to 2x200Gbps, 8x50G PAM4 lanes) with ultra-low latency and power. Backed by our lifetime warranty and rigorous testing—save 70-90% vs. genuine OEM while enabling flexible, high-density scaling. New, in stock at Resilient Tec, with free standard shipping on all US orders.
Key Features
- Massive Cost Savings & Flexibility: Affordable breakout for bridging 400G switches to 200G nodes—ideal for evolving AI/HPC setups.
- Plug-and-Play: Hot-pluggable connectors with pre-programmed EEPROM for seamless NVIDIA/Mellanox compatibility.
- Ultra-Low Power & Latency: Passive design (<0.5W total consumption) for dense, efficient racks.
- Lifetime Warranty: Fully tested in production environments.
- Use Cases: Fan-out in multi-node GPU clusters (e.g., H100 to legacy), four-node DGX Spark / GB10 clusters, spine-leaf expansions, or hybrid InfiniBand/Ethernet fabrics.
Technical Specifications
- Data Rate: 400Gbps (400GbE / NDR, split to 2x200Gbps HDR, 8x50Gbps PAM4)
- Connectors: 1x QSFP-DD to 2x QSFP56
- Lane Mapping: TX1-4 / RX1-4 to leg 1, TX5-8 / RX5-8 to leg 2 — each leg runs 200G CR4 (4 lanes)
- Length: 2 meters (6.6 ft)
- Cable Type: Passive Twinax Copper (typically 28-30AWG)
- Power Consumption: <0.5W
- Operating Temperature: 0°C to 70°C
- Standards Compliance: IEEE 802.3cd, SFF-8665, QSFP-DD/QSFP56 MSA, CMIS compliant
- Cross-reference: NVIDIA sells this cable under the full OPN MCP7H60-W002R26 (26 AWG, 2 m); the family is documented as MCP7H60-W0xxRyy. Also sold as the 1 m MCP7H60-W001, 1.5 m MCP7H60-W01AR30 and 3 m MCP7H60-W003.
Compatibility
Programmed for full compatibility with NVIDIA/Mellanox ecosystems and MSA-compliant platforms, including:
- NVIDIA Quantum-2 QM9700/QM9790 switches (400G ports)
- NVIDIA ConnectX-7 / ConnectX-6 adapters (200G ports)
- NVIDIA DGX H100/A100 systems and multi-node setups
- NVIDIA DGX Spark and other GB10 systems (ASUS Ascent GX10, Dell Pro Max GB10, HP ZGX Nano AI Station) — switch-side breakout, one port to two nodes
- MikroTik CRS804-4DDQ and CRS812-8DS-2DQ-2DDQ 400G QSFP-DD switches
- Other QSFP-DD/QSFP56 platforms (Arista 7800R, Cisco Nexus, etc.)
NVIDIA DGX Spark and GB10 Clustering
This is the switch-side cable for a four-node DGX Spark or GB10 cluster. One 400G QSFP-DD switch port fans out into two 200G QSFP56 legs, and each leg feeds one Spark at a full, unsplit 200G. The Spark's cage is QSFP112, which accepts a QSFP56 plug, and the Spark runs 200G as 4x50G PAM4 — exactly what each leg of this cable carries.
Compatible with MikroTik CRS804-4DDQ and CRS812 QSFP56-DD ports (set speed=200G-baseCR4 on lanes 1 and 5). Four DGX Sparks = 2 cables.
Breakout belongs on the switch end only. NVIDIA does not enable or validate port splitting on the DGX Spark's ConnectX-7, so a splitter at the Spark end will not give you two working links. Use this cable to fan a switch port out to two Sparks — never to fan a Spark out to two of anything.
When you need it, and when you don't. If your switch has a 400G port for every node — a MikroTik CRS804-4DDQ, for example — you can run one straight 200G QSFP56 DAC per Spark and skip breakout entirely. This cable is the right answer when the switch has fewer 400G ports than you have nodes (such as the CRS812-8DS-2DQ-2DDQ, with only two QSFP-DD ports), or when you want to keep switch ports free to grow past four nodes. Either way, set the switch port speed to 200G manually rather than relying on auto-negotiation, and plan on MTU 9000 with both logical halves of each Spark port driven in parallel — a Spark port that is only half-loaded tops out near 100Gbps.
Not sure whether your build needs breakout or straight cables? See the DGX Spark interconnect reference or send us your node count and switch model and we will confirm exactly what to buy before you order.
Need different lengths, 4x100G breakouts, or bulk quotes? Contact us—we specialize in high-bandwidth breakout solutions for AI workloads.
📋 COMPATIBILITY DOCUMENTATION
Download comprehensive compatibility report: MCP7H60-W002_Compatibility_Report
Guide includes: Pre-tested compatibility matrix for NVIDIA Quantum-2 QM9700/QM9790 switches, ConnectX-7 400G to ConnectX-6 200G breakout configurations, DGX Spark / GB10 switch-side breakout notes, EEPROM data and recognition protocols, QSFP-DD/QSFP56 MSA compliance guidelines, fan-out implementation instructions.
NVIDIA/Mellanox Compatible 400G QSFP-DD to 2x200G QSFP56 Passive Breakout Direct Attach Copper (DAC) Cable - 2 Meters
This third-party passive breakout DAC cable is fully compatible with NVIDIA/Mellanox LinkX series (e.g., MCP7H60-W002 equivalents), providing a cost-effective fan-out solution: one 400G QSFP-DD port to two 200G QSFP56 ports. Perfect for hybrid AI training clusters, HPC environments, and hyperscale data centers mixing 400G and 200G infrastructure.
Engineered to MSA standards, it supports full 400Gbps (split to 2x200Gbps, 8x50G PAM4 lanes) with ultra-low latency and power. Backed by our lifetime warranty and rigorous testing—save 70-90% vs. genuine OEM while enabling flexible, high-density scaling. New, in stock at Resilient Tec, with free standard shipping on all US orders.
Key Features
- Massive Cost Savings & Flexibility: Affordable breakout for bridging 400G switches to 200G nodes—ideal for evolving AI/HPC setups.
- Plug-and-Play: Hot-pluggable connectors with pre-programmed EEPROM for seamless NVIDIA/Mellanox compatibility.
- Ultra-Low Power & Latency: Passive design (<0.5W total consumption) for dense, efficient racks.
- Lifetime Warranty: Fully tested in production environments.
- Use Cases: Fan-out in multi-node GPU clusters (e.g., H100 to legacy), four-node DGX Spark / GB10 clusters, spine-leaf expansions, or hybrid InfiniBand/Ethernet fabrics.
Technical Specifications
- Data Rate: 400Gbps (400GbE / NDR, split to 2x200Gbps HDR, 8x50Gbps PAM4)
- Connectors: 1x QSFP-DD to 2x QSFP56
- Lane Mapping: TX1-4 / RX1-4 to leg 1, TX5-8 / RX5-8 to leg 2 — each leg runs 200G CR4 (4 lanes)
- Length: 2 meters (6.6 ft)
- Cable Type: Passive Twinax Copper (typically 28-30AWG)
- Power Consumption: <0.5W
- Operating Temperature: 0°C to 70°C
- Standards Compliance: IEEE 802.3cd, SFF-8665, QSFP-DD/QSFP56 MSA, CMIS compliant
- Cross-reference: NVIDIA sells this cable under the full OPN MCP7H60-W002R26 (26 AWG, 2 m); the family is documented as MCP7H60-W0xxRyy. Also sold as the 1 m MCP7H60-W001, 1.5 m MCP7H60-W01AR30 and 3 m MCP7H60-W003.
Compatibility
Programmed for full compatibility with NVIDIA/Mellanox ecosystems and MSA-compliant platforms, including:
- NVIDIA Quantum-2 QM9700/QM9790 switches (400G ports)
- NVIDIA ConnectX-7 / ConnectX-6 adapters (200G ports)
- NVIDIA DGX H100/A100 systems and multi-node setups
- NVIDIA DGX Spark and other GB10 systems (ASUS Ascent GX10, Dell Pro Max GB10, HP ZGX Nano AI Station) — switch-side breakout, one port to two nodes
- MikroTik CRS804-4DDQ and CRS812-8DS-2DQ-2DDQ 400G QSFP-DD switches
- Other QSFP-DD/QSFP56 platforms (Arista 7800R, Cisco Nexus, etc.)
NVIDIA DGX Spark and GB10 Clustering
This is the switch-side cable for a four-node DGX Spark or GB10 cluster. One 400G QSFP-DD switch port fans out into two 200G QSFP56 legs, and each leg feeds one Spark at a full, unsplit 200G. The Spark's cage is QSFP112, which accepts a QSFP56 plug, and the Spark runs 200G as 4x50G PAM4 — exactly what each leg of this cable carries.
Compatible with MikroTik CRS804-4DDQ and CRS812 QSFP56-DD ports (set speed=200G-baseCR4 on lanes 1 and 5). Four DGX Sparks = 2 cables.
Breakout belongs on the switch end only. NVIDIA does not enable or validate port splitting on the DGX Spark's ConnectX-7, so a splitter at the Spark end will not give you two working links. Use this cable to fan a switch port out to two Sparks — never to fan a Spark out to two of anything.
When you need it, and when you don't. If your switch has a 400G port for every node — a MikroTik CRS804-4DDQ, for example — you can run one straight 200G QSFP56 DAC per Spark and skip breakout entirely. This cable is the right answer when the switch has fewer 400G ports than you have nodes (such as the CRS812-8DS-2DQ-2DDQ, with only two QSFP-DD ports), or when you want to keep switch ports free to grow past four nodes. Either way, set the switch port speed to 200G manually rather than relying on auto-negotiation, and plan on MTU 9000 with both logical halves of each Spark port driven in parallel — a Spark port that is only half-loaded tops out near 100Gbps.
Not sure whether your build needs breakout or straight cables? See the DGX Spark interconnect reference or send us your node count and switch model and we will confirm exactly what to buy before you order.
Need different lengths, 4x100G breakouts, or bulk quotes? Contact us—we specialize in high-bandwidth breakout solutions for AI workloads.
📋 COMPATIBILITY DOCUMENTATION
Download comprehensive compatibility report: MCP7H60-W002_Compatibility_Report
Guide includes: Pre-tested compatibility matrix for NVIDIA Quantum-2 QM9700/QM9790 switches, ConnectX-7 400G to ConnectX-6 200G breakout configurations, DGX Spark / GB10 switch-side breakout notes, EEPROM data and recognition protocols, QSFP-DD/QSFP56 MSA compliance guidelines, fan-out implementation instructions.