NVIDIA Spectrum-X SN5610 Switch

The NVIDIA SN5610 is a next-generation Spectrum-X switch in the SN5000 series. NVIDIA Spectrum™-X SN5000 series switches are the fifth generation of Spectrum Ethernet switches and are purpose-built to accelerate data center fabrics. With port speeds ranging from 10GbE to 800GbE, the SN5000 series delivers accelerated Ethernet to every data center without compromising performance or features.

These switches are ideally suited for customers who deploy high-performance computing (HPC) clusters based on ProLiant DL, XL and Cray servers using Ethernet technology.

The SN5000 switch systems deliver high performance and consistent low latency while supporting advanced data center networking features, making them ideal for cloud networks and end-to-end data center fabrics. Powered by the NVIDIA Spectrum-4 ASIC, the SN5000 series provides dynamic, flexible shared buffers and predictable performance. As part of the Spectrum platform, SN5000 systems are pretested and prevalidated with NVIDIA’s full portfolio of Ethernet networking technologies, including NVIDIA® BlueField® data processing units (DPUs), ConnectX® SmartNICs, and LinkX® interconnects. Each element is fully compatible with standard Ethernet fabrics, while the end-to-end switch-to-host solution provides the performance and features required to accelerate cloud-native applications at scale.

The SN5610 smart-leaf/spine/super-spine offers 64 ports of 800GbE in a dense 2U form factor. The SN5600 is ideal for accelerated Ethernet deployments and enables both standard leaf/spine designs with top-of-rack switches as well as end-of-row (EoR) topologies. The SN5610 offers diverse connectivity in combinations of 1 to 800GbE and boasts an industry-leading total throughput of 51.2Tb/s.

Key Features

• – Up to 800GbE per port with 51.2 Tb/s of aggregate switching bandwidth
• – 64 OSFP ports in 2U form factor
• – 51.2 Tb/s non-blocking aggregate bandwidth
• – Supports standard leaf-and-spine designs with top-of-rack switches and end-of-row topologies
• – Quick and easy setup and management
• – Fully shared packet buffer provides a fair, predictable, and high-bandwidth data path
• – Ultra-low latency for efficient communication
• – Network operating system NVIDIA Cumulus Linux
• – Accelerated RDMA over Converged Ethernet (RoCE) for AI, cloud, storage, and high-performance computing (HPC)
• – RS-232 serial console port (1 × RJ-45) and USB 2.0 port (1 × USB)
• – Out-of-band (OOB) management interface: 1GbE, RJ-45

Switch Specifications

• – NVIDIA Spectrum-4-based SN5610 switch system
• – 64 800 Gb/s non-blocking ports
• – 128 400 Gb/s
• – 64 octal small form-factor pluggable (OSFP) connectors
• – Ethernet (800 Gbps) per port bidirectional
• – Switching capacity: 51.2 Tb/s
• – 80GB NVMe SSD
• – Thermal: 5 Fans. N+1 Redundant 60mm x 60mm
• – Supports passive copper or active copper or fiber cable with optical module
• – 4 AC hot-swappable power supplies
• – AMD EPYC 3251, 8-Cores, Secured boot

Management Ports (Managed Switch)

• – 1 × RJ-45 port (1 Gb/s Ethernet)
• – 1 × RJ-45 port (UART) and 1 × USB 3.0 port

Connectors and cabling

• – OSFP ports
• – Passive or active copper or optical module with fiber cables

Indicators

• – Per OSFP port status LED for link and activity
• – System status LEDs for the system, fans, and power supplies
• – Unit ID (beacon) LED

Power Supply

• – N+N Redundant, 5 FANS
• – Input range: 200-240Vac; US min: 2 phases of 100-110v – in total, at least 208v
• – 80 Gold+ and ENERGY STAR certified

Cooling

• – Connector to power airflow (reverse)
• – Hot-swappable fan units

NVIDIA Spectrum-X SN2201 Switch

The NVIDIA® Spectrum™ SN2000 series switches are the second generation of NVIDIA switches and are purpose-built for leaf, spine, and super-spine data center applications. The SN2000 series provides maximum flexibility, with port speeds ranging from 1GbE to 100GbE and port density that enables full-rack connectivity to servers at 1GbE, 20GbE, 25GbE, 40GbE, 50GbE, or 100GbE. In addition, the uplink ports support a variety of blocking ratios to meet application requirements.

The SN2000 series is ideal for building wire-speed, cloud-scale Layer 2 and Layer 3 networks. The SN2000 platforms deliver high performance and consistent low latency while supporting advanced software-defined networking features, making them an ideal choice for web-scale IT, cloud, hyperconverged storage, and data analytics applications.

The SN2201 is ideal as an out-of-band (OOB) management switch or as a top-of-rack (ToR) switch that connects up to 48 1GBASE-T host ports with nonblocking 100GbE spine uplinks. Featuring advanced hardware and software, ASIC-level telemetry, and a fully shared 16MB buffer, the SN2201 brings distinctive capabilities to 1GbE switching.

Key Features

• – Hardware-accelerated histograms track and summarize queue depths at sub-microsecond granularity
• – Inband Network Telemetry (INT)–ready hardware
• – Streaming telemetry
• – Up to 256,000 shared forwarding entries

NVIDIA NDR IB Quantum-2 QM9700 Switch

The NVIDIA^® Quantum-2 InfiniBand platform provides AI developers and scientific researchers with the highest networking performance available to take on the world’s most challenging problems. Innovative NVIDIA In-Network Computing acceleration engines provide ultra-low latency and double the data throughput, while delivering the scalability and feature-rich capabilities required for supercomputers, artificial intelligence, and hyperscale cloud data centers.

NVIDIA Quantum-2 enhances and extends In-Network Computing acceleration technology with preconfigured and programmable engines such as NVIDIA Scalable Hierarchical Aggregation and Reduction Protocol (SHARP) ^™, Message Passing Interface (MPI) tag matching, MPI_Alltoall, and programmable cores, as well as full transport offload with RDMA, GPUDirect RDMA, and GPUDirect Storage to deliver the best cost per node and return on investment (ROI).

The NVIDIA Quantum-2-based QM9700 and QM9790 switch systems deliver 64 ports of 400Gb/s InfiniBand per port, packed into a 1U standard chassis. Each switch carries an impressive 51.2 terabits per second (Tb/s) of aggregated bidirectional throughput and a landmark capacity exceeding 66.5 billion packets per second (BPPS). Supporting NVIDIA’s advanced 400Gb/s interconnect technology, NVIDIA Quantum-2 is a high-speed, extremely low-latency, and scalable networking solution. Key features incorporated include state-of-the-art technologies such as remote direct-memory access (RDMA), adaptive routing, and NVIDIA® Scalable Hierarchical Aggregation and Reduction Protocol (SHARP)™. Unlike other networking solutions, NVIDIA InfiniBand incorporates self-healing network capabilities, quality of service, enhanced virtual lane mapping, and advanced congestion control, maximizing overall application throughput. As ideal rack-mounted InfiniBand solutions, the QM9700 and QM9790 400Gb/s InfiniBand fixed-configuration switches offer exceptional flexibility, with support for various topologies like Fat Tree, SlimFly, DragonFly+, multi-dimensional Torus, and others.

They maintain backward compatibility with previous generations and support a broad software ecosystem. The QM9700 400Gb/s InfiniBand switches, enhancing NVIDIA In-Network Computing technologies, incorporate the third generation of NVIDIA SHARP technology, SHARPv3.

Key Features

• – 64 ports, each providing NDR (400Gbps) performance
• – 51.2 Tb/s non-blocking aggregate bandwidth
• – NDR200 Supports up to 128 ports of HDR (200 Gbps)
• – Quick and easy setup and management
• – Improved performance by removing fabric congestion
• – Backward compatibility with NVIDIA HDR InfiniBand technology for easier migration

NVIDIA NetQ Software for Spectrum-X Switches (On-Premises)

Network Operations with NetQ

NVIDIA NetQ™ is a highly scalable, modern network operations toolset that provides real-time visibility, troubleshooting, and validation for Cumulus fabrics. NetQ uses telemetry to deliver actionable insights into the health of your data center network, helping ensure that your AI network fabric operates smoothly.

How does NetQ work?

Data Collection

Using agents on the switches and hosts, NetQ is able to collect telemetry data across the entire network.

Data Processing

As a central control point, NetQ stores and processes information to provide actionable insights and complete visibility.

Data Visualization

The rich graphical user interface (GUI) simplifies operations and increases efficiency by quickly highlighting issues and alerts.

Further information available on https://www.nvidia.com/en-us/networking/ethernet-switching/netq/

Data Sheet: https://nvdam.widen.net/s/cqlbxvtw98/ethernet-switching-netq-datasheet

Key Features

• – Validations, Tracing, Flow telemetry analysis
• – RDMA over Converged Ethernet (RoCE) monitoring
• – Software upgrade management
• – Snapshot and compare
• – Topology
• – High-availability clustering
• – Precision Time Protocol (PTP) monitoring
• – Application-specific integrated circuit (ASIC) monitoring

Proof Points

• – Simplifies scaling of NVIDIA Cumulus® Linux
• – Speeds mean time to innocence
• – Reduces opex, downtime
• – Increases productivity
• – Maximizes value of network infrastructure

For the most up-to-date information on HPE Services, please refer to the HPE Services – Supplemental QuickSpecs, which provides a comprehensive and regularly updated overview of available services.

Have feedback on QuickSpecs? We’re listening