NVIDIA HGX-2

Achieve faster AI insight with NVIDIA HGX-2

Traditional data center architecture is no longer able to support the computational demands of today's modern AI workloads. As developers create increasingly large and accelerated computing clusters, the scale of data centers is reaching its limit. The NVIDIA HGX-2 multi-precision computing platform provides a solution for almost limitless AI computing that can achieve faster insights to transform the world.

As an industry-leading provider of customized server, storage and workstation solutions, Thinkmate provides the NVIDIA technology that data center architects require to transform their data centers with an NVIDIA HGX-2 platform to meet the requirements of AI and high-performance computing workloads.

The power of the NVIDIA HGX-2 platform

The NVIDIA HGX-2 cloud server platform offers multi-precision computing capabilities with unique flexibility to support the future of computing. The HGX-2 platform allows high-precision calculations using FP64 and FP32 for scientific simulations and computing, while also enabling FP16 and INT8 for AI training and inference. This versatility enables data centers to meet the requirements of the increasing number of applications that combine high-performance computing with AI.

The NVIDIA HGX-2 platform offers best-in-industry performance in the highly parallel computer architecture required for deep learning, machine learning and HPC. Users can rely on NVIDIA's complete solution stack of hardware and software to solve previously unsolvable problems. As the strongest compute node for data centers, HGX-2 replaces 300 CPU servers for AI training, 544 CPU servers for machine learning and accelerates HPC 156X faster than a CPU-only server.

NVIDIA provides partners with HGX-2 GPU baseboards, design guidelines, and access to GPU computing technologies, helping to rapidly advance AI cloud computing.

NVIDIA HGX-2 components from Thinkmate

Thinkmate has been providing customers in business, education, and government with custom workstations, servers, and storage solutions for more than 25 years. With highly reliable technology and a strong commitment to superior customer service, we are among the top white box server solutions providers in the industry.

Thinkmate offers the NVIDIA HGX-2 GPU baseboard along with the NVIDIA Tesla V100 GPUs and other technologies that power the HGX-2 platform. As the world's most advanced data center GPU ever built, the V100 is designed to dramatically accelerate deep learning, machine learning, high-performance computing and graphics. A single V100 Tensor Core GPU can provide the same performance of nearly 32 CPUs, making it the world's most powerful, scalable, and versatile computing platform.

Configuring NVIDIA technology at Thinkmate

Thinkmate's world-class configurator makes it easy to cost out and compare NVIDIA HGX-2 technology online. You can begin by choosing a base model server in a 1U-4U form factor with 1-10 GPUs, 4-24 drive bays, and 1-4 processors. Then you can select from a broad array of processors, drives, memory cards, network and storage cards, software, operating systems, and GPU accelerators, including the V100 as well as the NVIDIA P100, P40, and T4 server products.

In addition to NVIDIA HGX-2 baseboards and components, you'll find a wide variety of other NVIDIA components at Thinkmate, including an NVIDIA RTX server for a data science workstation, NVIDIA GeForce servers and NVIDIA Virtual Compute Servers. We also offer additional server and storage virtualization products, including VMware vSAN server products.

Why Thinkmate is your #1 source for HGX-2 technology

When you come to Thinkmate to transform your data center with NVIDIA HGX-2 technology, you can count on:

  • 100% customized solutions. Every Thinkmate machine is built to your exact specifications.
  • An award-winning 3-year warranty. A 3-year warranty with advanced parts replacement can be paired with all Thinkmate machines.
  • Superior customer service. You'll always be connected to a live person when you contact Thinkmate customer service for answers to questions and resolution of issues.
  • Highly competitive pricing. Our strong relationships with direct suppliers – the same people who supply the big-name manufacturers – enable us to offer prices that are on par with or lower than our competitors.
  • Equipment made in the USA. Every Thinkmate machine is built and stress-tested in our state-of-the-art facilities near Boston, Massachusetts.

FAQs: What is NVIDIA HGX-2?

What is NVIDIA HGX-2?

NVIDIA HGX-2 is a multi-precision computing platform with unprecedented versatility and unique flexibility to support the future demands of high-performance computing. NVIDIA's stack-to-stack enhancements across hardware, software, and libraries provide up to 10X faster training of advanced AI in six months. This revolutionary platform allows multi-precision calculations using FP64 and FP32 for scientific computing and simulations, while also enabling FP16 and INT8 for AI training and inference.

What are the benefits of NVIDIA HGX-2?

The HGX-2 platform replaces 300 CPU servers for AI training and accelerates HPC by 156 times over a CPU-only system. Thanks to NVIDIA NVSwitch powered by NVIDIA NVLink, the HGX-2 platform creates a unified networking fabric that allows the entire node to function as a single enormous GPU. Researchers can deploy models of unprecedented scale and solve the most complex HPC problems without being limited by compute capability.


Speak with an Expert Configurator at 1-800-371-1212