843400-B21 HPE Apollo 840Z InfiniBand EDR 100GB 2-Port Mezzanine FIO Adapter

Overview of HPE 843400-B21 Apollo InfiniBand 100Gb Adapter 

The HPE 843400-B21 Apollo InfiniBand EDR 100Gb 2-Port 840z Mezzanine FIO Adapter is a high-performance networking solution engineered for advanced data center environments, high-performance computing (HPC) clusters, and enterprise-grade server infrastructures. Designed to deliver ultra-low latency and exceptional throughput, this mezzanine form factor adapter enhances connectivity for next-generation workloads.

General Information

• Manufacturer: HPE 
• Part Number: 843400-B21
• Product Type: Network Adapter

Technical Specifications

• Device Type: Network Adapter
• Interface Type: PCI Express 3.0 x16 Mezzanine
• PCIE Revision: PCIE 3.0 compliant
• Ports: InfiniBand EDR x 2
• Data Link Protocol: InfiniBand
• Data Transfer Rate: Up to 100 Gbps
• Processor: 1 x Mellanox ConnectX-4
• Standards Compliance: IBTA 1.3

Compatible Systems

• HPE ProLiant XL230k Gen10 Compute Tray
• HPE ProLiant XL260a Gen9
• HPE ProLiant XL260a Gen9 Compute Tray

Supported Operating Systems

• SUSE Linux Enterprise Server 12 SP1
• CentOS 7.2 (64-bit)
• Red Hat Enterprise Linux 7.2 (64-bit)

The HPE 843400-B21 Apollo InfiniBand 100Gb 2-Port Adapter

The HPE 843400-B21 Apollo InfiniBand EDR 100Gb 2-Port 840Z Mezzanine FIO Adapter is a high-performance networking component engineered for extreme-scale computing environments where ultra-low latency, high bandwidth, and deterministic data throughput are critical. Designed and optimized by this adapter plays a central role in modern high-performance computing (HPC), artificial intelligence (AI) clusters, scientific simulations, and large-scale enterprise data centers that rely on InfiniBand interconnect technology for scalable distributed processing.

High-Performance InfiniBand Architecture and EDR Technology

Enhanced Data Rate (EDR) InfiniBand is a next-generation networking standard that provides up to 100 gigabits per second per port. Unlike conventional Ethernet-based networking, InfiniBand is purpose-built for HPC workloads, offering extremely low latency, hardware-based transport offloading, and remote direct memory access (RDMA) capabilities. The 843400-B21 adapter leverages these features to ensure near-zero CPU overhead during data transmission, allowing compute resources to remain fully dedicated to application workloads rather than network processing.

Dual-Port 100Gb Connectivity Advantage

The dual-port design of the 840Z Mezzanine FIO Adapter allows simultaneous communication across two independent InfiniBand links. This enables advanced network topologies such as fat-tree, hypercube, or dragonfly configurations, which are commonly deployed in supercomputing environments. Each port can be configured for redundancy or load balancing depending on cluster design requirements. By distributing data traffic across two 100Gb links, the adapter reduces congestion hotspots and ensures stable performance during peak computational workloads such as molecular modeling, weather forecasting, and deep neural network training.

Mezzanine Form Factor and Apollo System Integration

The mezzanine architecture of the HPE 843400-B21 adapter is specifically designed for tight integration into Apollo server systems. Rather than occupying standard PCIE slots, the adapter connects through a dedicated mezzanine interface that provides direct access to system buses while maintaining compact internal layouts. This design enables higher compute density per rack unit, making it ideal for data centers that prioritize scalability and space efficiency. The reduced cabling complexity also enhances airflow management, resulting in improved thermal performance and reduced cooling overhead.

Remote Direct Memory Access (RDMA)

One of the most critical features of InfiniBand EDR technology is RDMA, which allows one system to access the memory of another system without involving its CPU. The 843400-B21 adapter fully supports RDMA operations, ensuring that data transfers occur with minimal latency and CPU intervention. This capability is essential for scaling distributed applications efficiently, as it reduces context switching overhead and significantly improves throughput consistency under heavy workloads.

Topology Flexibility and Fabric Design

InfiniBand networks support multiple topology designs, and this adapter provides the flexibility required to implement them efficiently. Whether deployed in a fat-tree architecture for balanced bandwidth distribution or a dragonfly topology for reduced hop counts, the adapter ensures consistent data flow performance.

Signal Integrity and High-Speed Transmission

At 100Gb per port, maintaining signal integrity becomes a critical engineering challenge. The adapter incorporates advanced PCB design techniques, impedance matching, and high-quality transceiver components to ensure stable data transmission over InfiniBand links. These design optimizations reduce packet loss, minimize retransmissions, and ensure efficient utilization of available bandwidth in high-density environments.

High-Speed Networking

This InfiniBand adapter is built to support demanding workloads that require fast and reliable data transfer. With dual-port 100Gbps EDR capability, it ensures scalable and efficient communication between compute nodes, storage systems, and distributed applications.

Mellanox ConnectX-4 Processor Integration

At the core of the adapter lies a powerful Mellanox ConnectX-4 processor, which plays a critical role in managing network traffic, reducing CPU overhead, and optimizing packet processing efficiency.