NVIDIA HGX B100
Advance Generative AI with NVIDIA Blackwell
Purpose-built to drive the next wave of generative AI innovation, the NVIDIA HGX B100 delivers unparalleled computing power to scale complex AI models. Experience the power of the future today with NVIDIA Blackwell B100.
Unrivalled Performance in...
Acceleration
Configuration of 8 GPUs, up to 112 petaFLOPs of AI performance
Memory
Up to 1.5TB of fast memory for speedups
GPU Bandwidth
Up to 192 GB HBM3e massive memory bandwidth
Networking
5th-gen NVLink with 1.8TB/s of GPU-to-GPU interconnect and InfiniBand networking
Boost LLM with NVIDIA HGX B100
Built with Blackwell Architecture
Introducing the Blackwell-architecture packed with 208 billion transistors and manufactured using TSMC's cutting-edge 4NP process, these GPUs redefine performance. Featuring two interconnected GPU dies forming a unified chip with a 10 TB/second link, the Blackwell GPUs are a true force to be reckoned with. The second-generation Transformer Engine and 4-bit floating-point AI inference support double the compute and model sizes.
Peak Security and Reliablity
The innovative RAS Engine ensures unparalleled reliability, availability, and serviceability, while its AI-based preventative maintenance capabilities maximise system uptime and reduce operating costs. Advanced confidential computing features safeguard your AI models and customer data without sacrificing performance.
Benefits of NVIDIA HGX B100
A New Class of AI Superchip
Custom-built 4NP TSMC, this GPU pushes the boundaries of performance. Its an innovative design consisting of two GPU dies, each intricately connected by 10 TB/second chip-to-chip link. This seamless integration forms a unified GPU to handle the most demanding tasks
LLM Innovation
With cutting-edge Blackwell Tensor Core technology, combined with TensorRT-LLM and Nemo Framework innovations, get unmatched acceleration for LLM and Mixture-of-Experts models
AI Capabilities
Second-generation Transformer Engine introduces groundbreaking innovations, including 8-bit floating-point (FP8) precision and support for new numerical formats. Multi-GPU scaling with the fifth-generation NVLink interconnect with 1.8TB/s of bidirectional GPU-to-GPU bandwidth
Reliable AI
The RAS Engine incorporates AI-based preventative maintenance capabilities directly at the chip level for advanced diagnostics and forecasting of potential reliability issues before they occur. It boosts system uptime and maintain seamless operations for your large-scale AI deployments
Data Security
Advanced confidential computing capabilitieS to secure AI models and customer data without compromising performance. New native interface encryption protocols for data-sensitive industries
Data Acceleration
Accelerate database queries and experience the highest performance in data analytics and data science with a dedicated decompression engine. Supporting the latest formats, this engine allows you to tackle vast amounts of data
AI Innovation
FP4 precision with 14 petaFLOPS delivers a massive leap in AI performance to lead in innovation.
Ultra-fast NVLink
The NVIDIA HGX B100 integrates powerful Blackwell B100 GPUs with ultra-fast NVLink interconnects for accelerated computing and generative AI workloads.
Massive Memory
192 GB HBM3e of memory to handle the largest AI models and datasets for seamless scalability.
New Wave of Gen AI with the NVIDIA HGX B100
