NVIDIA MSB7800-ES2F 100Gb/s InfiniBand Switch 36-Port 7.2Tb/s HPC & AI Switch

Model: MSB7800-ES2F | P2C Airflow | 36 x QSFP28 | 1U Top-of-Rack

The NVIDIA MSB7800-ES2F is a 100Gb/s InfiniBand fixed switch delivering 36 ports in a compact 1U chassis. Designed for small to extremely large clusters, it provides top-of-rack leaf connectivity with non‑blocking performance. This model features power‑to‑core (P2C) airflow, ideal for data center cold aisle configurations. Enterprises and research institutions leverage the SB7800 series to handle high‑performance computing, AI training, and storage workloads without compromising latency or scalability. The integrated x86 dual-core CPU runs the feature-rich MLNX-OS®, enabling full chassis management, while advanced telemetry and SHARP acceleration transform how collective communications are processed across the fabric.

Each port delivers line-rate 100Gb/s InfiniBand, supporting full bisection bandwidth and non‑blocking architecture. Aggregate throughput reaches 7.2 Tb/s.

NVIDIA Scalable Hierarchical Aggregation and Reduction Protocol offloads collective operations (MPI_Allreduce, etc.) from the host CPU to the switch data path, drastically reducing data traversals and latency.

80 Plus Gold certified, ATIS weighted power as low as 122W. Redundant (1+1) hot‑swappable power supplies and fan modules maximize uptime and serviceability.

On‑board Intel Celeron 1047UE processor, 4GB DDR3, 16GB SSD — runs full switch OS, enabling robust telemetry, remote management, and automation.

Integrate with UFM® (Unified Fabric Manager) to gain AI‑powered analytics, real‑time telemetry, predictive troubleshooting and proactive data center orchestration.

MSB7800-ES2F provides power‑to‑core (P2C) airflow; ideal for hot‑aisle containment. This aligns with standard data center cooling patterns.

Traditional MPI operations require multiple passes of data between endpoints, consuming CPU cycles and increasing job completion time. NVIDIA SHARP technology integrates collective acceleration directly into the switch fabric. By performing arithmetic reductions (sum, min, max, etc.) as data passes through the aggregation nodes, the SB7800 reduces the number of packets traversing the network by up to a factor of 4× for certain workloads. This breakthrough not only cuts MPI operation time dramatically but also liberates valuable CPU resources for actual computation. Combined with Remote Direct Memory Access (RDMA) over Converged Ethernet (RoCE) support and native InfiniBand, the MSB7800-ES2F excels in HPC, deep learning, and large‑scale simulations.

Founded in 2008, Starsurge is a technology-driven provider of network hardware, IT services, and system integration solutions. With a global customer base spanning government, healthcare, manufacturing, education, finance, and enterprise sectors, we deliver reliable networking infrastructure including switches, NICs, wireless controllers, cabling, and IoT solutions. Our experienced technical and sales teams ensure responsive service, custom development, and multilingual support. By partnering with Starsurge, clients build scalable, future-proof networks with a customer-first approach.

For volume quotes, custom firmware imaging, or extended warranty on NVIDIA MSB7800 series — contact our InfiniBand specialists today.