This site requires Javascript in order to view all its content. Please enable Javascript in order to access all the functionality of this web site. Here are the instructions how to enable JavaScript in your web browser.

QM8700 InfiniBand Series

Quantum HDR 200Gb/s InfiniBand Smart Edge Switches

Faster servers, high-performance storage, and increasingly complex computational applications are driving data bandwidth requirements to new heights. NVIDIA QM8700 InfiniBand switches provide extremely high networking performance by delivering up to 16Tb/s of non-blocking bandwidth with extremely low latency. Static routing, adaptive routing, and advanced congestion management optimize computing efficiencies, making QM8700 ideal for top-of-rack leaf connectivity or for small to extremely large clusters.

High Bandwidth

High-performance computing (HPC) and AI environments need every last bit of the bandwidth delivered by NVIDIA high data rate (HDR) 200Gb/s switch systems. NVIDIA QM8700 switches provide up to 40 ports of 200Gb/s full bi-directional bandwidth per port. And novel NVIDIA HDR100 technology supports up to 80 ports of 100Gb/s, enabling HDR switches to provide double-density radix for 100Gb/s data speeds, reducing the cost of network design and network topologies.

LEARN MORE ›

Highlights

LINK SPEED

200Gb/s

NUMBER OF PORTS

40HDR/80^*HDR100

MAX. THROUGHPUT

16Tb/s

SWITCH SIZE

Highlights

LINK SPEED

200Gb/s

NUMBER OF PORTS

40HDR/80^*HDR100

MAX. THROUGHPUT

16Tb/s

SWITCH SIZE

POWER CONSUMPTION (ATIS)

274W

*When used in conjunction with NVIDIA^® ConnectX^®-6 adapter cards, the QM8700 can achieve up to 80 ports of HDR100 100Gb/s

World-Class InfiniBand Performance

In-Network Computing

NVIDIA Scalable Hierarchical Aggregation and Reduction Protocol (SHARP)™ In-Network Computing offloads collective communication operations from the CPU to the switch network by aggregating large data vectors at wire speed—crucial for machine learning applications.

LEARN MORE ›

Self-Healing Networking

NVIDIA InfiniBand with self-healing networking capabilities, overcomes link failures and achieves network recovery 5,000X faster than any software-based solution—enhancing system performance, scalability, and network utilization.

LEARN MORE ›

NVIDIA Mellanox Unified Fabric Management (UFM)

UFM Management

NVIDIA Unified Fabric Management (UFM^®) platforms combine enhanced, real-time network telemetry with AI-powered cyber intelligence and analytics; to realize higher utilization of fabric resources and a competitive advantage, while reducing OPEX.

LEARN MORE ›

Key Features

40 HDR 200Gb/s ports in a 1U switch
80 HDR100 100Gb/s ports in a 1U switch
16Tb/s aggregate switch throughput
Up to 15.8 billion messages per second
Enables offloading of collective operations from the CPU to the switch network
Quality-of-service enforcement

1+1 power supply
Makes the fabric capable with self-healing autonomy
Adaptive routing
Collective offloads
Virtual lane (VL) mapping (VL2VL)

Benefits

Industry-leading switch platform in performance, power, and density
Collective communication acceleration
Quick and easy setup and management
Maximizes performance by removing fabric congestions
Backwards compatible to previous speeds

QM8700 Series Comparison Table

	Link Speed	Ports	Height	Switching Capacity	Cooling System	Interface	Number of PSUs	Management	Subnet Manager
QM8700	200Gb/s	40	1U	16Tb/s	Air cooled	QSFP56	2	Inband/outband	+
QM8790	200Gb/s	40	1U	16Tb/s	Air cooled	QSFP56	2	Inband	-