NVIDIA Expands vGPU Software to Accelerate Workstations, AI Compute Workloads

Designers, engineers, researchers, creative professionals all need the flexibility to run complex workflows – no matter where they’re working from.

With the newest release of NVIDIA virtual GPU (vGPU) technology, enterprises can provide their employees with more power and flexibility through GPU-accelerated virtual machines from the data center or cloud.

Available now, the latest version of our vGPU software brings GPU virtualization to a broad range of workloads — such as virtual desktop infrastructure, high-performance graphics, data analytics and AI —  thanks to its support for the new NVIDIA A40  and NVIDIA A100 80GB GPUs. The new release also supports the NVIDIA GPU Operator, a software framework that simplifies GPU deployment and management.

Powerful Performance for Power Users

NVIDIA RTX Virtual Workstation (vWS) software is a major component of the vGPU portfolio, designed to help users run graphics-intensive applications on virtual workstations. With NVIDIA A40 powering NVIDIA RTX vWS, professionals can achieve up to 60 percent⁽¹⁾ faster virtual workstation performance per user and 2x⁽²⁾ faster rendering than the previous generation RTX 6000 GPUs.

NVIDIA A40 includes second-generation RT Cores and third-generation Tensor Cores to help users accelerate workloads like photorealistic rendering of movie content, architectural design evaluations, and virtual prototyping of product designs. With 48GB of GPU memory, professionals can easily work with massive datasets and run workloads like data science or simulation with even larger model sizes.

NVIDIA A40 support with the latest vGPU software enables complex graphics workloads to be run in a virtualized environment with performance that is on par with bare metal.

“With support for NVIDIA’s latest vGPU software, and the new NVIDIA A40 with Citrix Hypervisor 8.2 and Citrix Virtual Desktops, we can continue providing the performance customers need to run graphics-intensive visualization applications as their data and workloads grow,” said Calvin Hsu, vice president of product management at Citrix. “The combination of Citrix and NVIDIA virtualization technologies provides access to these applications from anywhere, with an experience that is indistinguishable from a physical workstation.”

The NVIDIA vGPU January 2021 software release supports the NVIDIA A100 80GB to deliver increased memory bandwidth, unlocking more power for large models. This builds on the September release, which introduced compute features that included support for the NVIDIA A100 Tensor Core GPU, the most advanced GPU for AI and high performance computing.

Additional new features include simplified GPU management in Kubernetes through NVIDIA GPU Operator, which is now supported with NVIDIA Virtual Compute Server and NVIDIA RTX vWS software. Containers, including the GPU-optimized software available in the NGC catalog, can be easily deployed and managed in VMs.

With this new release, customers and IT professionals can continue managing their multi-tenant workflows running in virtual machines using popular hypervisors, like Red Hat Enterprise Linux, while the certified GPU Operator brings a similar experience to containerized deployments on top of Red Hat virtualization platforms using Red Hat OpenShift.  

“The combination of NVIDIA’s latest generation A40 GPU and NVIDIA vGPU software, supported with Red Hat Enterprise Linux and Red Hat Virtualization, offers a powerful platform capable of serving some of the most demanding workloads ranging from AI/ML to visualization in the oil and gas as well as media and entertainment industries,” said Steve Gordon, director of product management at Red Hat. “As organizations transform and increasingly use containers orchestrated by Kubernetes as key building blocks for their applications, we see Red Hat OpenShift as a likely destination for containerized and virtualized workloads alike.”

To find a certified server, see the NVIDIA vGPU Certified Server page.

Learn more about NVIDIA vGPU software portfolio, which includes:

• NVIDIA RTX Virtual Workstation (RTX vWS) (formerly known as Quadro Virtual Data Center Workstation or Quadro vDWS)
• NVIDIA Virtual Compute Server (vCS)
• NVIDIA Virtual PC (vPC) (formerly known as GRID vPC)
• NVIDIA Virtual Applications (vApps) (formerly known as GRID vApps)

_{1. Tested on a server with Intel Xeon Gold 6154 3.0GHz 3.7GHz Turbo, RHEL 8.2, vGPU 12.0 software, running four concurrent users per GPU, RTX6000P-6Q versus A40-12Q, running SPECviewperf 2020 Subtest 4K 3dsmax-07 composite.}

_{2. Iray 2020.1. Render time (seconds) of NVIDIA Endeavor scene.}