Brand: NVIDIA | Category: Network Adapters & NICs
SKU: NVID-MCX85346ABASNEAT | Part #: MCX85346ABAS-NEAT | MPN: MCX85346ABAS-NEAT
Contact for Pricing — Request a Quote
The NVIDIA ConnectX-8 SuperNIC 400GbE Adapter from NVIDIA is enterprise-grade Network Adapters & NICs hardware built for data centers. NVIDIA ConnectX-8 SuperNIC delivers 400GbE line-rate throughput with in-network compute acceleration purpose-built for AI fabric and GPU-dense enterprise cluster networking. Available in brand new condition. Available from Omnixon Global — contact our team to request a quote.
| Brand | NVIDIA |
| Category | Network Adapters & NICs |
| SKU | NVID-MCX85346ABASNEAT |
| Part Number | MCX85346ABAS-NEAT |
| Condition | New |
| Product Line | ConnectX-8 SuperNIC |
| Form Factor | PCIe Low-Profile / Full-Height Dual-Slot |
| Host Interface | PCIe Gen 5.0 x16 |
| Network Ports | 2 x QSFP112 (400GbE per port) or 1 x 400GbE dual-port configuration |
| Maximum Bandwidth | 400 Gbps full-duplex per port (800 Gbps aggregate) |
| Ethernet Standards | 400GBASE-SR8, 400GBASE-DR4, 400GBASE-LR4, 100GbE, 50GbE, 25GbE, 10GbE backward compatible |
| RDMA Protocol | RoCEv2 (RDMA over Converged Ethernet v2) |
| In-Network Compute | NVIDIA SHARP v4 — hardware AllReduce, AllGather, Broadcast, Barrier offload |
| Congestion Control | Hardware-based Explicit Congestion Notification (ECN), Dynamic Rate Limiting, Adaptive Routing telemetry export |
| Virtualization | SR-IOV with up to 256 Virtual Functions (VFs) per adapter; VXLAN, GENEVE, NVGRE hardware offload |
| Security Acceleration | Hardware-accelerated TLS 1.3, IPsec inline encryption/decryption, AES-256-GCM at line rate |
| Time Synchronization | IEEE 1588v2 Precision Time Protocol (PTP), hardware timestamping |
| Message Rate | Up to 300 million messages per second |
| Latency | Sub-1 microsecond port-to-port RDMA latency |
| CPU Offload Engines | Stateless offloads: LRO, LSO, checksum; DPDK PMD support; GPUDirect RDMA |
| GPUDirect Support | NVIDIA GPUDirect RDMA and GPUDirect Storage (GDS) — peer-to-peer GPU memory access without CPU staging |
| Operating System Support | Linux (RHEL, Ubuntu, SLES), Windows Server 2022, VMware ESXi 8.x |
| Management | NVIDIA MLNX_OFED driver stack; OpenFabrics Alliance (OFA); Telemetry via gRPC/Prometheus; NVIDIA UFM integration |
| Power Consumption | ≤ 75W (PCIe slot powered); auxiliary power connector for configurations requiring higher TDP |
| Operating Temperature | 0°C to 55°C ambient |
| Certifications | CE, FCC Class A, RoHS 3 compliant, TAA compliant |
| Launch | Q1 2025 |
The NVIDIA ConnectX-8 SuperNIC 400GbE Adapter is enterprise networking hardware for data-center, campus, and branch deployments. Common roles include core/distribution switching, server uplinks, top-of-rack fabric, and edge connectivity in mixed Cisco/Juniper/Arista environments.
Key specifications for the NVIDIA ConnectX-8 SuperNIC 400GbE Adapter: new condition; manufacturer NVIDIA; product line ConnectX-8 SuperNIC; form factor PCIe Low-Profile / Full-Height Dual-Slot; host interface PCIe Gen 5.0 x16; network ports 2 x QSFP112 (400GbE per port) or 1 x 400GbE dual-port configuration; maximum bandwidth 400 Gbps full-duplex per port (800 Gbps aggregate). Manufacturer part number MCX85346ABAS-NEAT. For the full datasheet with electrical, environmental, and compliance details, contact our pre-sales engineering team.
The NVIDIA ConnectX-8 SuperNIC 400GbE Adapter interoperates with standard switching protocols (LACP, MLAG/VPC, 802.1Q, OSPF/BGP). For specific cross-vendor scenarios — Cisco-to-Arista MLAG peering, fabric integration with Cumulus/SONiC, or migration from legacy Catalyst — our network engineers map a compatibility plan as part of the RFQ.