NVIDIA GB200 NVL72

The Future of Computing with NVIDIA GB200 NVL72

Unleash next-level performance with the groundbreaking NVIDIA GB200 compute tray—engineered around NVIDIA’s flexible MGX architecture. Featuring 2 high-performance Grace CPUs and 4 powerful Blackwell GPUs, it’s built to handle the toughest generative AI, data analytics, and high-performance computing (HPC) workloads with unmatched efficiency.

NVIDIA GB200 NVL72 Server Racks

Unrivalled Performance in...

LLM Inference

1.7 TB fast memory for trillion-parameter LLMs

Multimodal Transformer Models

900 GB/s bandwidth accelerates multimodal transformer tasks

Generative 3D Capabilities

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

Data Processing

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

Computing Revolution with Blackwell GB200

Real-Time Inference for Trillion-Parameter LLMs

Unlock real-time inference at an entirely new scale with the GB200 NVL72. Powered by the second-generation Transformer Engine, FP4 precision, and fifth-generation NVIDIA NVLink, it delivers up to 30x faster performance for trillion-parameter LLMs. Enhanced Tensor Cores introduce advanced microscaling formats for greater throughput and precision. With NVLink interconnects and liquid cooling, the 72-GPU rack eliminates communication bottlenecks and maximises inference speed.

Massive LLM Training at High Speed

Supercharge LLM training with the GB200 NVL72’s second-gen Transformer Engine and FP8 precision, achieving up to 4x faster results. Fifth-gen NVLink enables scaling up to 576 GPUs within a single domain, delivering over 1 petabyte per second bandwidth and access to 240 TB of high-speed memory. Integrated InfiniBand networking ensures lightning-fast communication and seamless performance at scale.

Benefits of NVIDIA Blackwell GB200

Revolutionary Blackwell Architecture

The NVIDIA Blackwell architecture marks a new era in accelerated computing, delivering unmatched performance, energy efficiency, and scalability. Built with 208 billion transistors using TSMC’s custom 4NP process, it's engineered for breakthrough innovation.

Eco-Efficient Performance

Step into sustainable computing with the liquid-cooled GB200 NVL72. Achieve 25x the performance of air-cooled H100 systems—at the same power level—while significantly lowering your energy use and environmental impact.

Next-Level Data Acceleration

Transform your data workflows with high-bandwidth memory, NVLink-C2C, and built-in decompression engines. Speed up critical database queries by 18x over traditional CPUs and enjoy a 5x improvement in total cost of ownership.

Next-Gen CPU Power

The NVIDIA Grace CPU revolutionises data centre computing with exceptional speed and memory performance. Delivering 2x the energy efficiency of today’s top server CPUs, it powers AI, cloud, and HPC workloads with ease.

Ultra-Fast Interconnect

Fifth-generation NVIDIA NVLink unlocks the scale and speed needed for exascale computing and trillion-parameter models. It enables ultra-fast, low-latency GPU-to-GPU communication across your entire cluster.

Scalable High-Speed Networking

NVIDIA Quantum-X800 InfiniBand, Spectrum-X800 Ethernet, and BlueField-3 DPUs work in unison to scale performance across massive GPU deployments—ensuring high-speed, reliable networking for the most demanding applications.

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

Reserve your NVIDIA GB200 NVL72 today!

NVIDIA GB200 NVL72 Processor Chip

Frequently Asked Questions

We build our services around you. Our product support and product development go hand in hand to deliver you the best solutions available.

The GB200 is powered by NVIDIA’s groundbreaking Blackwell architecture, designed for next-generation performance.

With ultra-fast memory bandwidth, NVLink interconnects, and dedicated decompression engines, the GB200 can accelerate critical database queries up to 18x compared to standard CPUs—delivering up to 5x greater cost-efficiency for data processing tasks.

The 5th generation NVLink enables up to 130 TB/s bandwidth for seamless multi-GPU communication. Combined with NVIDIA’s high-speed InfiniBand, Ethernet, and DPU technologies, this architecture allows efficient scaling across thousands of Blackwell GPUs.

The GB200 features 192 GB of high-performance memory, enabling intensive workloads to run with ease.

The NVIDIA GB200 can draw up to 1,200 watts of power under maximum load.

Pricing for the GB200 NVL72 is available upon reservation. Contact us to learn more and secure your unit.