NVIDIA MCX653106A-HDAT ConnectX-6 200G HDR InfiniBand Smart Network Interface Card
Product Details:
| Brand Name: | Mellanox |
| Model Number: | MCX653106A-HDAT-SP |
| Document: | connectx-6-infiniband.pdf |
Payment & Shipping Terms:
| Minimum Order Quantity: | 1pcs |
|---|---|
| Price: | Negotiate |
| Packaging Details: | outer box |
| Delivery Time: | Based on inventory |
| Payment Terms: | T/T |
| Supply Ability: | Supply by project/batch |
|
Detail Information |
|||
| Products Status: | Stock | Application: | Server |
|---|---|---|---|
| Condition: | New And Original | Type: | Wired |
| Max Speed: | Up To 200 Gb/s | Ethernet Connector: | QSFP56 |
| Model: | MCX653106A-HDAT | Name: | MCX653106A-HDAT-SP Mellanox Network Card 200gbe High Speed Smart Safe |
| Highlight: | Mellanox ConnectX-6 network card,200G HDR InfiniBand NIC,Smart network interface card with warranty |
||
Product Description
The NVIDIA® ConnectX®-6 MCX653106A-HDAT is a high-performance dual-port Smart network card designed for the most demanding data center environments. Part of the NVIDIA Quantum InfiniBand platform, this adapter delivers up to 200Gb/s per port (HDR InfiniBand) with groundbreaking In-Network Computing acceleration. It significantly reduces CPU overhead, slashes latency, and boosts application performance for High-Performance Computing (HPC), Artificial Intelligence (AI), and hyperscale cloud workloads. This advanced nic card represents a pivotal component for building efficient, scalable, and intelligent fabric infrastructures.
- Ultra-High Speed: Dual-port 200Gb/s HDR InfiniBand (200G per port) connectivity
- In-Network Computing: Offloads computation and aggregation to the network, freeing host CPU resources and improving scalability
- Hardware Security: Integrated XTS-AES 256/512-bit block-level encryption, FIPS compliant, offloaded from the CPU
- Extreme Performance: Achieves extremely low latency and a high message rate of up to 215 million messages/second
- Advanced Offloads: Comprehensive RDMA, TCP/IP, NVMe over Fabrics (NVMe-of), and Switch ASAP² for OVS offload
- Robust Connectivity: Supports both PCIe 3.0 and 4.0 host interfaces, ensuring compatibility with modern server platforms
- Flexible Deployment: Available in standard PCIe stand-up form factor with QSFP56 cages for flexible cable choices
This network card is built on the innovative ConnectX-6 architecture, leveraging several key technologies:
- IBTA 1.3 Compliant InfiniBand: Delivers native RDMA, adaptive routing, and advanced congestion control
- PCI Express Gen 4.0: Maximizes host interface bandwidth to fully utilize the 200Gb/s network speed
- NVIDIA In-Network Computing Engines: Enable collective operation offloads and in-network memory access, reducing inter-node communication bottlenecks
- Hardware-Based I/O Virtualization (SR-IOV): Supports up to 1000 virtual functions for secure, high-performance multi-tenant environments
- Enhanced Ethernet RoCE Support: Provides high-performance RDMA over converged Ethernet networks
The MCX653106A-HDAT acts as an intelligent co-processor for server networking. Instead of burdening the main CPUs, critical networking functions are offloaded to the adapter's specialized hardware:
- Data Path Acceleration: RDMA operations allow data to move directly between application memory and the network, bypassing the OS kernel
- Compute Offload: In-Network Computing engines perform data aggregation and collective operations within the switch fabric
- Security Processing: AES-XTS encryption/decryption of data-at-rest happens at line rate on the nic card itself
- Protocol Processing: TCP/IP, NVMe-of, and storage protocol headers are processed in hardware, minimizing latency
This architecture transforms the network card from a passive data pipe into an active participant in the computing workflow.
- HPC & Supercomputing Clusters: For running large-scale simulations and modeling with MPI-based applications requiring ultra-low latency and high message rates
- AI/ML Training Farms: Accelerates deep learning training by providing high-bandwidth, low-latency connectivity between GPU servers, reducing job completion times
- Hyperscale Cloud & Data Centers: Enhances efficiency for storage disaggregation (NVMe-of), virtualized networks, and data-intensive analytics platforms
- High-Frequency Trading & Financial Modeling: Where microsecond latencies and deterministic performance are critical
- Consolidated Workloads: Ideal for servers running a mix of HPC, AI, and storage workloads, thanks to its versatile offload capabilities
| Attribute | Specification |
|---|---|
| Model / Product Alias | MCX653106A-HDAT (related to mq9700 series) |
| Form Factor | PCIe Stand-Up, Full-Height |
| Network Ports | 2 x QSFP56 |
| Data Rate per Port | Up to 200 Gb/s (HDR InfiniBand) |
| Host Interface | PCIe 3.0/4.0 x16 |
| Key Offloads | RDMA, In-Network Computing, NVMe-of, TLS/XTS-AES Encryption |
| Virtualization | SR-IOV (Up to 1000 VFs) |
| Supported OS | Linux (RHEL, SLES, Ubuntu), Windows Server, VMware |
Why choose the MCX653106A-HDAT over other network card solutions?
- Beyond Connectivity: It's not just a nic card; it's an In-Network Computing accelerator that improves overall system efficiency and scalability
- Superior Efficiency: Dramatically reduces CPU utilization for networking and storage tasks, allowing CPUs to focus on core applications
- Unmatched Performance: Delivers industry-leading latency and message rates, crucial for HPC and AI workloads
- Integrated Security: Hardware-based encryption provides high-performance security without compromising speed, a feature not found in many standard adapters
- Future-Proof Design: With support for PCIe 4.0 and the latest InfiniBand standards, it protects infrastructure investments
We provide comprehensive support for every MCX653106A-HDAT adapter sold:
- Full Warranty: Backed by standard manufacturer warranty
- Reliable Supply: Stable inventory ensures timely delivery to meet project deadlines
- Technical Expertise: Our engineering team offers pre- and post-sales consultation for seamless integration into your HPC, AI, or cloud environment
- Long-Term Partnership: We provide ongoing support for firmware updates, compatibility checks, and optimization guidance
- Server Compatibility: Verify that your server has an available PCIe x16 slot (preferably Gen4) and adequate cooling/power capacity
- Driver & Firmware: Always use the latest NVIDIA-approved drivers and firmware from the official support site for optimal stability and performance
- Thermal Management: Ensure proper airflow within the server chassis, as high-performance adapters generate heat. Do not block the card's heatsink
- Electrostatic Discharge (ESD): Handle the network card with proper ESD precautions. Install it while the server is powered off and disconnected
- Software Environment: Confirm operating system compatibility. For advanced features like In-Network Computing, specific software libraries (e.g., HPC-X, NCCL) may be required
We are a leading global supplier of enterprise networking hardware with over ten years of industry experience. Our operations are supported by a large-scale facility and a robust technical team, allowing us to serve a vast and diverse customer base.
As an authorized distributor for premier brands such as NVIDIA Mellanox, Ruckus, Aruba, and Extreme, we specialize in providing original, brand-new equipment. Our extensive portfolio includes network switches, network cards (like the ConnectX series), wireless access points, controllers, and cabling solutions.







