<img alt="" src="https://secure.insightful-enterprise-intelligence.com/783141.png" style="display:none;">

Access NVIDIA H100 in minutes from just $2.06/hour. Reserve here

Deploy 8 to 16,384 NVIDIA H100 SXM GPUs on the AI Supercloud. Learn More

NVIDIA GB200 NVL72

New Wave of Computing with NVIDIA GB200 NVL72

Introducing the cutting-edge NVIDIA GB200 compute tray, a revolutionary solution utilising the full potential of NVIDIA's innovative MGX design. With 2 powerful Grace CPUs and 4 advanced Blackwell GPUs, it delivers unparalleled performance to tackle the most demanding generative AI, data analytics and HPC tasks.

nvidia-gb200-nvl72

Unrivalled Performance in...

nvidia gb200 nvl72 llm inference

LLM Inference

1.7 TB fast memory for trillion-parameter LLMs

nvidia-gb200-nvl72-multimodal-transformer-models

Multimodal Transformer Models

900 GB/s bandwidth accelerates multimodal transformer tasks

nvidia-gb200-nvl72-generative-3d-capabilities

Generative 3D Capabilities

1.8 TB/s bidirectional throughput per GPU for generative models for 3D data

NVIDIA GB200 NVL72 Data Processing

Data Processing

Experience faster data processing speed of up to 18x over typical CPUs

Computing Revolution with NVIDIA GB200 NVL72 

nvidia-gb200-nvl72-real-time-inference

Real-Time Inference for Trillion-Parameter LLMs

Experience unparalleled real-time LLM inference with the GB200 NVL72. Its cutting-edge second-generation Transformer Engine, FP4 AI and fifth-generation NVIDIA NVLink deliver a stunning 30x faster performance for trillion-parametre language models. This breakthrough is powered by a new generation of Tensor Cores, introducing advanced microscaling formats for superior accuracy and throughput. The GB200 NVL72 leverages NVLink and liquid cooling to create a massive 72-GPU rack to overcome communication bottlenecks

massive-LLM-training

Massive LLM Training at High Speed

Accelerate massive-scale training with the GB200 NVL72's second-generation Transformer Engine featuring FP8 precision. A remarkable 4X faster training for LLM, redefining the boundaries of possibility. This groundbreaking performance is powered by fifth-generation NVLink, which connects up to 576 GPUs in a single NVLink domain with over 1 PB/s total bandwidth and 240 TB of fast memory. The GPU is als equipped with InfiBand networking for super speed and efficiency.

Benefits of NVIDIA Blackwell GB200

groundbreaking-blackwell-architecture

Groundbreaking Blackwell Architecture

The NVIDIA Blackwell architecture leads groundbreaking advancements for accelerated computing into a new era of unparalleled performance, efficiency and scalability. It packs 208 billion transistors and is manufactured using a custom-built TSMC 4NP process.

sustainable-computing

Sustainable Computing

Embrace a greener future with the liquid-cooled GB200 NVL72 racks. Reduce your carbon footprint and energy consumption while delivering 25X more performance than NVIDIA H100 air-cooled infrastructure at the same power.

accelerated-data-processing

Accelerated Data Processing

Experience high-bandwidth memory performance, NVLink-C2C, and dedicated decompression engines to accelerate key database queries by 18X compared to CPUs. Enjoy a 5X better TCO and revolutionise your data processing capabilities.

breakthrough-CPU-performance

Breakthrough CPU Performance

The NVIDIA Grace CPU redefines modern Data Centre computing with outstanding performance and memory bandwidth. Experience 2X the energy efficiency of today's leading server processors while powering AI, cloud, and HPC applications with unprecedented speed.

seamless-interconnectivity

Seamless Interconnectivity

Get the full potential of exascale computing and trillion-parametre AI models with the fifth-generation NVIDIA NVLink. This scale-up interconnect unleashes accelerated performance for swift and seamless communication between every GPU in your server cluster.

high-performance-networking

High-Performance Networking

NVIDIA Quantum-X800 InfiniBand, NVIDIA Spectrum X800 Ethernet, and NVIDIA BlueField®-3 DPUs enable efficient scalability across hundreds and thousands of Blackwell GPUs for optimal application performance.

Technical Specifications

GPU: NVIDIA GB200 NVL72

GPU Memory: 192 GB HBM3e

Power: 1200W

FP4 Tensor Core
20 petaFLOPS
FP8/FP6 Tensor Core
10 petaFLOPS
INT8 Tensor Core
10 petaOPS
FP16/BF16 Tensor Core
5 petaFLOPS
TF32 Tensor Core
2.5 petaFLOPS
FP64 Tensor Core
45 teraFLOPS
GPU memory
Up to 192 GB HBM3e
Bandwidth
Up to 8 TB/s
Multi-Instance GPU (MIG)
7
Decompression Engine
Yes
Decoders
2x 7 NVDEC, 2x 7 NVJPEG
Power
Configurable up to 1,200W
Interconnect
5th Generation NVLink: 1.8TB/s, PCIe Gen6: 256GB/s
FP4 Tensor Core
20 petaFLOPS
FP8/FP6 Tensor Core
10 petaFLOPS
INT8 Tensor Core
10 petaOPS
FP16/BF16 Tensor Core
5 petaFLOPS
TF32 Tensor Core
2.5 petaFLOPS
FP64 Tensor Core
45 teraFLOPS
GPU memory
Up to 192 GB HBM3e
Bandwidth
Up to 8 TB/s
Multi-Instance GPU (MIG)
7
Decompression Engine
Yes
Decoders
2x 7 NVDEC, 2x 7 NVJPEG
Power
Configurable up to 1,200W
Interconnect
5th Generation NVLink: 1.8TB/s, PCIe Gen6: 256GB/s

Reserve your NVIDIA GB200 NVL72  today!

reserve-your-nvidia-gb200-nvl72--today

Frequently Asked Questions

Our product support and product development go hand in hand to deliver you the best solutions available.

What are the key benefits of NVIDIA Blackwell GB200 for LLM inference and training?

The NVIDIA GB200 card delivers groundbreaking performance for LLM inference and training. Its second-gen Transformer Engine with FP4/FP8 precision, massive GPU interconnect bandwidth, and large HBM3e memory capacities enable real-time trillion-parameter LLM capabilities.

What advantages does the NVIDIA GB200 GPU Card offer for data processing workloads?

The NVIDIA GB200 GPU Card can accelerate key database queries by up to 18x over typical CPUs thanks to its high memory bandwidth, NVLink interconnects, and dedicated decompression engines. This enables a 5x better total cost of ownership for data processing workloads.

What interconnect technologies enable large-scale Blackwell GPU deployments?

The 5th gen NVIDIA NVLink GPU interconnect with up to 130 TB/s bandwidth allows seamless multi-GPU communication. Combined with NVIDIA's high-speed InfiniBand, Ethernet, and DPU technologies, this provides efficient scalability across thousands of Blackwell GPUs.

How much is NVIDIA GB200 GPU Card memory?

The NVIDIA GB200 GPU Card has 192 GB memory capacity.

How much power does the NVIDIA GB200 GPU consume?

The NVIDIA GB200 power consumption is up to 1,200W.