InfiniBand 1.8 Spec Brings 200Gb/s Per Lane Speed, Scalability for RDMA Fabrics

In Focus: Infiniband

🌐 InfiniBand High-Performance Networking Fabric for AI and HPC Clusters
InfiniBand is a high-performance networking technology used to connect servers, GPUs, storage systems and accelerators in AI training clusters, supercomputers and high-performance computing environments. Known for low latency, high throughput and native RDMA support, InfiniBand remains a major fabric for tightly coupled compute workloads.
Why It Matters	Large AI training systems require fast, predictable communication across thousands of GPUs. InfiniBand provides a mature, purpose-built fabric for scale-out compute, helping reduce communication overhead and improve accelerator utilization in demanding AI and HPC deployments.
Category	Networking Technology
Typical Deployment	AI training clusters • HPC supercomputers • GPU fabrics • Research computing • Large-scale simulation
Key Capabilities	Native RDMA • Low latency • High throughput • Adaptive routing • Congestion control • Collective communication support • Scale-out GPU networking
Topics Covered	AI Infrastructure • GPU Networking • High-Performance Computing • RDMA • Supercomputing • Scale-Out AI • Data Center Fabrics
Key Companies	Dell Technologies • HPE • Lenovo • NVIDIA • Supermicro
Related Technologies	RDMA • RoCE • Ethernet • Ultra Ethernet • AI Infrastructure
Standards & Industry Groups	InfiniBand Trade Association

InfiniBand 1.8 Spec Brings 200Gb/s Per Lane Speed, Scalability for RDMA Fabrics

2 years ago

The InfiniBand Trade Association (IBTA) has announced significant advancements in its Volume 1 Release 1.8 Specification, which introduces critical enhancements ...

650 Group: AI Networking Surges over 100% YoY

3 years ago

The Data Center AI Networking market is expected to grow to over $3 billion in 2023, according to a newly ...

Tech Update: Ethernet vs Infiniband for GPU clusters

3 years ago

Wondering whether to use Infiniband or Ethernet for building large-scale clustered GPU environments? https://youtu.be/I2NKlVG3Ai0 Ram Velaga, SVP & GM, Switching ...

Meta’s next AI supercomputer to use NVIDIA 200 Gb/s InfiniBand fabric

4 years ago

Meta is building an AI research supercomputer that will leverage NVIDIA systems, InfiniBand networking and AI software to enable optimization ...

NVIDIA unveils Quantum-2 networking platform – a 400Gbps InfiniBand Switch

5 years ago

NVIDIA announced the next generation of its InfiniBand networking platform for cloud computing providers and supercomputing centers.NVIDIA Quantum-2 is a ...

NVIDIA intros Mellanox 400G InfiniBand – the 7th generation

6 years ago

NVIDIA introduced the seventh generation of Mellanox InfiniBand delivering ultra-low latency and double data throughput with NDR 400Gb/s, as well ...

Mellanox supplies 200G HDR InfiniBand to ECMWF

7 years ago

Mellanox Technologies will supply its 200 Gigabit HDR InfiniBand to the European Centre for Medium-Range Weather Forecasts (ECMWF) to accelerate ...

Mellanox extends Infiniband to 40km

7 years ago

Mellanox Technologies introduced its Quantum LongReach series of long-distance InfiniBand switches for connecting remote InfiniBand data centers together, or to ...

Mellanox supplies 200G InfiniBand for Lenovo’s liquid cooled servers

7 years ago

Mellanox Technologies has begun shipping liquid cooled HDR 200G Multi-Host InfiniBand adapters for the Lenovo ThinkSystem SD650 server platform, which ...