NVIDIA ConnectX-8 C8180 HHHL PCIe Gen 6.0 800Gb/s OSFP SuperNIC

NVIDIA® ConnectX®-8 SuperNIC™

The NVIDIA® ConnectX®-8 SuperNIC™ is optimised to supercharge hyperscale AI computing workloads. With support for both InfiniBand and Ethernet networking atup to 800 gigabits per second (Gb/s), ConnectX-8 SuperNIC delivers extremely fast, efficient network connectivity, significantly enhancing system performance for AI factories and cloud data center environments.

Colossal 800 Gb/s OSFP Transfers via PCIe 6.0 x16

Central to NVIDIA’s AI networking portfolio, ConnectX-8 SuperNICs fuel the nextwave of innovation in forming accelerated, massive-scale AI fabrics. They seamlessly integrate with next-gen NVIDIA networking platforms, providing up to 800 Gb/s of end-to-end connectivity. These platforms offer the robustness, feature sets, and scalability required for trillion-parameter GPU computing, AI data platforms, and agentic AI applications. With enhanced power efficiency, ConnectX-8 SuperNICs support the creation of increasingly sustainable AI data centers operating hundreds of thousands of GPUs, ensuring a future-ready infrastructure for AI advancements.

Stay Ahead in Super Intelligent Inference

ConnectX-8 SuperNICs enable advanced routing and telemetry-based congestion control capabilities, achieving the highest network performance and peak AI workload efficiency. Additionally, ConnectX-8 InfiniBand SuperNICs extend the capabilities of NVIDIA® Scalable Hierarchical Aggregation and Reduction Protocol (SHARP)™ to boost in-network computing in high-performance computing environments, further enhancing overall efficiency and performance for both training and inferencing at scale.

• 800 Gb/s HH/HL PCIe 6.0 x16 AI networking adapter for hyperscale AI data centres, GPU clusters and high-performance computing (HPC).
• InfiniBand and Ethernet connectivity with up to 800 Gb/s total bandwidth.
• PCIe Gen6 host interface with up to 48 lanes.
• InfiniBand speeds of 800/400/200/100 Gb/s; Ethernet speeds of 400/200/100/50/25 Gb/s.
• Single-port and dual-port configurations, with Ethernet support for up to eight split ports.
• NVIDIA Multi-Host support for up to four hosts.
• RDMA, RoCEv2, GPUDirect RDMA and GPUDirect Storage acceleration.
• Advanced programmable congestion control, routing, telemetry and high-speed packet reordering.
• In-network computing, NVIDIA SHARP, MPI acceleration and collective operation offloads for distributed AI training and inference.
• Ethernet offloads for IP/TCP/UDP checksums, LSO, LRO, GRO, RSS, TSS and SR-IOV.
• Accelerated switching and packet processing, VXLAN/GENEVE/NVGRE offloads, connection tracking, NAT and hierarchical QoS.
• PCIe HHHL single-port OSFP, dual-port QSFP112 and OCP 3.0 TSFF single-port OSFP form factors; dual ConnectX-8 mezzanine option.
• Hardware security features including secure boot, root of trust, encrypted flash, firmware protection, device attestation and inline IPsec/MACsec/PSP acceleration.
• Precision timing and synchronisation with IEEE 1588v2 PTP, SyncE, PTM, hardware timestamping and configurable PPS.
• Network boot support for InfiniBand, Ethernet, PXE, iSCSI and UEFI.
• AI/HPC software support for NCCL, HPC-X, DOCA UCC/UCX, Open MPI and MVAPICH2.
• IBTA v1.7 compliance; up to 16 million I/O channels and 2 GB messages.