NVIDIA Virtual Compute Server

Power intensive workloads with NVIDIA Virtual Compute Server

The majority of server workloads for midsize to large enterprises are now run on virtual machines (VMs). However, workloads for AI, data science, deep learning, and high-performance computing – which require GPU-accelerated servers – are the exception, as virtualization has been focused primarily on CPU-only servers. As a result, workloads requiring GPU servers are often isolated from other servers, limiting utilization and flexibility.

NVIDIA Virtual Compute Server software now enables hypervisor-based server virtualization, allowing data centers to power any compute-intensive workloads GPUs in a virtual machine. As you source the technology to run an NVIDIA Virtual Compute Server, you'll find a wide range of GPU-accelerated servers and NVIDIA processors at Thinkmate, a leading provider of custom-built machines for customers in business, education, and government.

NVIDIA is transforming virtualized compute

NVIDIA Virtual Compute Server enables data centers to accelerate server virtualization using the latest NVIDIA data center GPUs. With products like NVIDIA A100 Tensor Core GPUs, NVIDIA Virtual Compute Server makes it possible for virtual machines to run the most compute-intensive workloads for artificial intelligence, deep learning and data science.

Key features of NVIDIA Virtual Compute Server include:

  • GPU sharing. Only NVIDIA vGPU technology makes fractional GPU sharing possible. With NVIDIA, multiple VMs can share a GPU to maximize utilization for lighter workloads requiring GPU acceleration.
  • Management and monitoring. NVIDIA Virtual Compute Server provides support for monitoring at the app, guest and host level. Proactive management features enable live migration as well as suspend and resume functions.
  • GPU aggregation. Enable a VM to access more than one GPU for certain compute-intensive workloads. NVIDIA Virtual Compute Server supports multi-vGPU (where GPUs aren't directly connected) and peer-to-peer computing (where GPUs are connected through NVLink for higher bandwidth).
  • NVIDIA GPU Cloud (NGC). NGC is a hub for GPU-optimized software that streamlines workflows for machine learning, deep learning and HPC. NGC now supports virtualized environments with NVIDIA Virtual Compute Server.
  • ECC & page retirement. For applications that are sensitive to data corruption, error code correction (ECC) and page retirement provide higher reliability and are especially important in large-scale cluster-computing environments.
  • Peer-to-peer computing. NVIDIA virtual GPU technology now supports NVIDIA NVLink, a high-speed direct GPU-to-GPU interconnect that delivers higher bandwidth, improved scalability, and more links for multi-GPU configurations.
  • Multi-instance GPU. This revolutionary technology enables each NVIDIA A100 Tensor Core GPU to be partitioned into up to seven instances, running a VM on each instance to take advantage of the management, monitoring, and operational benefits of hypervisor-based server virtualization.

Virtual GPU technology from Thinkmate

Thinkmate has been a world-class provider of custom server, storage, and workstation solutions for more than 25 years. With deep technical knowledge and superior customer service, we enable our customers to source the custom-configured machines they need at highly competitive prices.

We offer a broad range of GPU-optimized servers featuring NVIDIA Tesla GPUs for AI, deep learning and data science. Our GPU servers generate massively parallel processing power and unrivaled networking flexibility, offering two double-width GPUs or up to five expansion slots in a 1U to deliver performance and quality optimized for the most computationally intensive applications. Thinkmate GPU servers include features that optimize cooling for even the most demanding applications, providing the ideal technology platform for GPU supercomputing solutions. We offer NVIDIA Virtual Compute Server software on many of our machines.

Additional NVIDIA products available through Thinkmate include servers featuring Tesla V100, P100 and T4 GPUs, along with options for a Quadro RTX server. You can also build a VMware server and other virtualized products through Thinkmate.

Configuring your virtual GPU server

To build and price out a GPU server with NVIDIA Virtual Compute Server software, start by selecting a base model from our collection of GPU-accelerated servers. You can select a base model based on its platform, available processors, maximum RAM, number of drives, form factor, and expansion possibilities. Then take your pick from a long list of components, adding items or swapping them out to configure your ideal machine. The price of each component is marked clearly in the configurator so you can easily compare apples to apples. You'll see a running total cost of your configured unit at the top of the page, so you can keep an eye on the price of your machine as you choose your options.

When you've completed your build, you can order it, request an official quote or submit it for review by our knowledgeable computer technicians.

Why Thinkmate is your best choice for an NVIDIA server

When you choose an NVIDIA Virtual Compute Server from Thinkmate, you'll be impressed by the many benefits our technology offers.

  • Exceptional reliability. Our dedication to building a highly reliable machine has made us a #1 solutions provider.
  • Completely customized machines. We don't do off-the-shelf technology – every Thinkmate server and storage solution is completely customized to the exact specifications of our customers.
  • Outstanding customer service. Every member of our team is dedicated to doing everything in their power to deliver the excellence you expect from Thinkmate technology, and you'll always be able to speak with a live person by phone or chat when you need answers to questions, help with configuration or resolution of issues.
  • USA-made machines. Thinkmate technology is built at our state-of-the-art facilities near Boston.
  • Competitive prices. We are able to offer exceptionally competitive pricing on our machines, thanks to our relationships with direct suppliers.

FAQs: What is NVIDIA Virtual Compute Server?

What is NVIDIA Virtual Compute Server?

NVIDIA Virtual Compute Server software enables data centers to accelerate server virtualization with the latest NVIDIA data center GPUs, running the most compute-intensive workloads on virtual machines.

What is NVIDIA Virtual Compute Server vs. GRID VPC/V apps vs. Quadro vDWS?

NVIDIA Virtual Compute Server is designed for compute-intensive server workloads such as deep learning, AI, and data science. GRID vPC and GRID vApps are client compute products for virtual graphics and are designed for use by creative or technical professionals and knowledge workers. NVIDIA Quadro vDWS enables users to use Quadro graphics applications on any device, anywhere.


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