900-9X760-0078-MB0 - NVIDIA 900-9X760-0078-MB0 200GbE Dual QSFP112 OCP3.0 Network Adapter

Name: NVIDIA 900-9X760-0078-MB0 200GbE Dual QSFP112 OCP3.0 Network Adapter
Brand: NVIDIA
SKU: 900-9X760-0078-MB0
Price: 2218 USD
Availability: InStock

900-9X760-0078-MB0 - NVIDIA 900-9X760-0078-MB0 200GbE Dual QSFP112 OCP3.0 Network Adapter | US IT Hardware

Product overview

The NVIDIA ConnectX-7 OCP3.0 SFF Adapter Card (P/N: 900-9X760-0078-MB0, OPN: MCX753436MS-HEAB) provides dual-port QSFP112 connectivity supporting 200GbE Ethernet and InfiniBand NDR200/HDR protocols over PCIe Gen5 x16 interface. Designed for hyperscale data centers, AI/ML clusters, HPC environments, and high-performance storage, this small form factor card delivers ultra-low latency, high throughput, and advanced offloads including ASAP² packet processing, GPUDirect Storage, SR-IOV virtualization, and TLS/IPsec/MACsec crypto acceleration. The thumbscrew pull-tab bracket enables easy installation in dense rack servers. Backward compatible with PCIe Gen4/3, it supports auto-negotiation for InfiniBand rates (NDR200, HDR, EDR, FDR, SDR) and Ethernet speeds (200/100/50/25/10GbE). Target workloads include cloud-native applications, distributed training, NVMe-oF storage fabrics, and secure multi-tenant environments. Compliance with PCIe Gen5, IEEE 802.3, IBTA v1.5, and RoHS ensures enterprise-grade reliability. Typical power consumption is 24.5W with passive cables, powered via 12V_edge and +3.3V_edge rails.

Dual QSFP112 200GbE/NDR200.
PCIe 5.0 x16 OCP3.0 SFF form factor.
ASAP2+GPUDirect+SR-IOV offloads.
24.5W typical power consumption.
IBTA v1.5 + IEEE 802.3 compliant.
Secure Boot + Crypto acceleration.

Expedited and overnight shipping options are available at checkout, priced by destination and service.
Purchase orders accepted from qualified institutions.
Refunds within 14 days and replacements within 30 days of delivery with an RMA (25% restocking fee on customer-initiated returns), and every part carries our Standard 30-Day Warranty - see our return policy or contact US IT Hardware.

Featuring

Manufacturer part #	900-9X760-0078-MB0
Brand	NVIDIA
Condition	New
Offloads	ASAP², GPUDirect Storage, SR-IOV, VirtIO, TLS/IPsec/MACsec
Compliance	PCIe Gen5, IEEE 802.3, IBTA v1.5, RoHS
Device Type	Network Adapter - Dual-Port QSFP112 - OCP3.0 SFF

Network & Communication

Data Rates	IB: NDR200/HDR/EDR/FDR/SDR ¦ ETH: 200/100/50/25/10GbE

Security

Crypto

Compatible systems

Optimized for PCIe Gen5 OCP3.0 servers from Dell, HPE, Lenovo with MLNX_OFED drivers across Linux distros and VMware.

Dell PowerEdge
Model: XE9680
Form factor: 4U Rack
Generation: PCIe Gen5
OCP 3.0 NIC slot required
HPE ProLiant
Model: DL380
Form factor: 2U Rack
Generation: Gen11
PCIe Gen5 OCP support
Lenovo ThinkSystem
Model: SR675 V3
Form factor: 2U Rack
Generation: Gen5
OCP 3.0 slot validated

Best for

AI/ML Training
Distributed deep learning with multi-node GPU clusters using NCCL, Horovod, and GPUDirect RDMA for tensor data movement in TensorFlow/PyTorch frameworks.
High-Performance Computing
HPC simulations and MPI workloads requiring NDR200 InfiniBand with SHARP RoCEv2 offloads for weather modeling, genomics, and CFD in OpenMPI environments.
Hyperscale Cloud
Virtualized cloud infrastructure with SR-IOV and DPDK acceleration for NFV, container orchestration in Kubernetes, and east-west traffic in OpenStack.

Details

Shipping, warranty & returns

Customer reviews

No reviews yet — be the first to review this part.

Secure Boot	Supported

Voltage	12V_edge, +3.3V_edge
Power Consumption	Typical 24.5W passive cables ¦ Up to 17W/port QSFP112

Dimensions	115.00 mm × 76.00 mm x 15.1 mm height
Form Factor	Small Form Factor (SFF) OCP3.0 ¦ 115x76mm
Bracket Type	Thumbscrew (Pull Tab)

Ports	2 x QSFP112 ¦ InfiniBand NDR200/HDR ¦ Ethernet 200GbE
Interface	PCIe 5.0 x16 ¦ Backward compatible Gen4/3 ¦ 32GT/s SerDes

Frequently asked questions

What PCIe generations does this adapter support?

Supports PCIe Gen5 x16 at 32GT/s as primary interface with full backward compatibility to Gen4 (16GT/s) and Gen3 systems. Auto-negotiation ensures optimal performance across server generations while maintaining PCIe 5.0 lane bifurcation and power management compliance.

What protocols and speeds are supported on the QSFP112 ports?

Dual QSFP112 ports support InfiniBand NDR200/HDR/EDR/FDR/SDR and Ethernet 200/100/50/25/10GbE with automatic lane negotiation. IBTA v1.5 compliance enables flexible 1x/2x/4x configurations ideal for both fat-tree HPC clusters and Ethernet spine-leaf data center fabrics.

What advanced offload features reduce CPU utilization?

Includes NVIDIA ASAP² packet processing, GPUDirect Storage/RDMA for GPU-NIC direct data transfer, SR-IOV with 256 virtual functions, and hardware crypto for IPsec/TLS/MACsec. These offloads free 90%+ of CPU cycles previously spent on networking in AI training, NVMe-oF, and NFV workloads.

Is this adapter suitable for OCP server platforms?

Yes, specifically designed as OCP NIC 3.0 SFF form factor (115x76mm) with thumbscrew pull-tab bracket for tool-less installation. Fully compliant with OCP 3.0 electrical, mechanical, and thermal specifications for seamless integration in hyperscale servers from OEMs supporting the OCP standard.

What is the typical power consumption profile?

Typical power draw is 24.5W with passive copper cables in PCIe Gen5 x16 mode. Each QSFP112 port supports up to 17W thermal budget for active optics or high-power transceivers while maintaining overall card power under 12V_edge and 3.3V_edge rails with advanced power gating.

Does it support virtualization and multi-tenancy?

Comprehensive virtualization support includes SR-IOV with 256 VFs per port, VirtIO-user with hardware acceleration, and VF representor for container networking. Secure Boot and MACsec encryption enable zero-trust multi-tenant deployments in VMware vSphere, KVM, and Kubernetes environments.

NVIDIA 900-9X760-0078-MB0 200GbE Dual QSFP112 OCP3.0 Network Adapter