NVIDIA BlueField-2 DPU - FHHL & HHHL Form Factor, 2xSFP56 / 2xQSFP56, 25/100GbE, Ultra-Low Latency for AI, HPC & Cloud

NVIDIA BlueField-2 Ethernet DPU is a high-performance, low-latency data processing unit designed for AI, HPC, and cloud infrastructures. It provides dual-port 25/100GbE connectivity via SFP56 or QSFP56, supporting deterministic networking, hardware-accelerated storage, and security offloads. With FHHL and HHHL form factors, BlueField-2 offloads CPU-intensive networking, storage, and security tasks, delivering predictable throughput and minimal jitter for GPU-dense servers and data center clusters. Its programmable DOCA datapath, advanced telemetry, and GPUDirect support enable efficient GPU-to-GPU communication, NVMe storage acceleration, and scalable cloud-native deployments. BlueField-2 ensures secure, high-bandwidth, and low-latency networking for modern AI, HPC, and enterprise workloads.

KEY FEATURES:

• High-speed connectivity: Dual 25/100GbE ports via SFP56 or QSFP56 for high-bandwidth networking
  
• Ultra-low latency: Optimised for deterministic GPU-to-GPU and server-to-server communication
  
• GPUDirect support: RDMA and Storage for direct data movement with minimal CPU involvement
  
• Hardware offload: Reduces CPU load for networking, storage, and security operations
  
• Programmable datapath: DOCA-programmable for SDN, NFV, and custom network functions
  
• Congestion control & traffic management: Adaptive routing, zero-loss RoCE, and hierarchical QoS
  
• Scalable design: FHHL & HHHL form factors for high-density AI, HPC, and cloud deployments
  
• Advanced telemetry: Real-time monitoring for ports, traffic, congestion, and system health
  
• Security acceleration: Secure boot, flash encryption, AES-GCM/AES-XTS, MACsec/IPsec, Cerberus root-of-trust
  
• Storage acceleration: NVMe-oF, NVMe/TCP, decompression, erasure coding/RAID, SNAP block storage offload
  
• Timing & synchronisation: IEEE 1588v2, SyncE, PPS In/Out, time-triggered scheduling for telecom and edge
  

PRODUCT SPECIFICATION:

APPLICATIONS:

AI & HPC Fabric Acceleration:

The BlueField-2 Ethernet DPU is engineered for AI, HPC, and GPU-accelerated clusters where ultra-low latency and deterministic networking are essential. With hardware offload for RoCEv2, congestion control, and advanced packet scheduling, it ensures high-throughput, reliable communication between GPUs and servers. GPUDirect RDMA support enables direct GPU-to-GPU data movement across nodes, bypassing the CPU to reduce latency and accelerate distributed AI training workloads. Ideal for GPU-dense servers, AI training clusters, and high-performance computing environments demanding predictable network performance.

Cloud Computing & Virtualised Workloads:

BlueField-2 enhances cloud infrastructure with hardware-accelerated networking for high I/O virtualised environments. Its programmable pipeline offloads VXLAN, NVGRE, and GENEVE encapsulation, improving virtual networking efficiency. Enhanced SR-IOV, multi-queue offloads, and queue scheduling reduce CPU load, freeing cores for tenant workloads. This ensures low-latency, scalable, and stable multi-tenant cloud deployments, supporting serverless computing, microservices, and high-density virtualisation.

Data Center Storage & NVMe Workloads:

Optimised for storage networking, BlueField-2 accelerates NVMe-over-Fabrics (NVMe-oF) and RDMA traffic, enabling near-local latency for disaggregated storage nodes. Zero-copy transfers and congestion-aware routing reduce CPU involvement, increasing efficiency for all-flash arrays, software-defined storage, and cloud-native databases. The DPU ensures predictable throughput, faster replication, and lower latency for mission-critical storage workloads.

Enterprise Networking & Secure Data Path Acceleration:

BlueField-2 delivers hardware-accelerated data path operations with lightweight security offloads. Line-rate encryption, access control, and secure transport protect enterprise workloads without impacting performance. Deterministic traffic control allows isolation of applications, enforcement of QoS policies, and consistent network efficiency during mixed traffic scenarios. Ideal for latency-sensitive enterprise workloads such as financial analytics, simulation, and real-time decision engines.

Telecom, 5G Core & Edge Infrastructure:

Telecom and edge deployments benefit from BlueField-2’s precise timing support and congestion-aware transport. Features like hardware timestamping, IEEE 1588v2, and RoCE congestion control enable low-latency, deterministic packet handling for 5G UPF, ORAN, and edge gateway applications. Scalable performance ensures stable operation in metro PoPs, edge data centres, and telecom clouds handling high volumes of small packets with strict timing requirements.

ADVANCED FEATURES:

• High-performance Data Path Offload: Offloads networking, storage, and security functions from the CPU, reducing jitter and freeing cores for application workloads.
  
• Ultra-low Latency Networking: Optimised for deterministic GPU-to-GPU and server-to-server communication across HPC and AI clusters.
  
• Programmable Datapath (DOCA): Supports SDN, NFV, and custom network functions via NVIDIA DOCA SDK.
  
• GPUDirect Support: RDMA and Storage enable direct GPU-to-GPU and GPU-to-storage data movement with minimal CPU involvement.
  
• Advanced Telemetry & Monitoring: Real-time metrics for ports, traffic, congestion, and health monitoring.
  
• Hardware Security Offloads: Secure boot, flash encryption, AES-GCM / AES-XTS, MACsec/IPsec, and Cerberus-compliant root-of-trust.
  
• Congestion Control & Traffic Management: Adaptive routing, zero-loss RoCE, and hierarchical QoS for predictable throughput.
  
• Storage Acceleration: NVMe-oF, NVMe/TCP, decompression, erasure coding, RAID, and SNAP block storage offload.
  
• Timing & Synchronisation: IEEE 1588v2, SyncE, PPS In/Out, and time-triggered scheduling for telecom and edge deployments.
  
• Flexible Port Configurations: Supports SFP56 (25/50GbE) and QSFP56 (100GbE) with multi-lane link aggregation and failover options.
  
• Form Factor Flexibility: FHHL / HHHL modules with M.2 / U.2 support for direct-attached storage, enabling high-density deployments.

PRODUCT DETAILS & ORDERING INFORMATION:

WARRANTY & SUPPORT:

Manufacturers are responsible for warranties, which are based on the purchase date and validity. We make sure the products we sell are delivered on time and in their original condition at Grabnpay. Whether it's physical damage or operational issues, our team will help you find a solution. We help you install, configure, and manage your devices. We'll help you file a warranty claim and guide you every step of the way.

All listed prices reflect standard market rates. Final commercial terms, including volume-based discounts, will be provided based on purchase volume, MOQ, and other applicable criteria. For more information about switches & other accessories, contact our technical support team.