Scaling the AI-Ready Data Center with NVIDIA RTX PRO 4500 Blackwell Server Edition and NVIDIA vGPU 20

AI integration is redefining mainstream enterprise applications, from productivity software like Microsoft Office to more complex design and engineering tools. This shift requires the modern data center to move beyond single-purpose silos.

For developers, gaining access to dedicated GPU compute can often be a bottleneck. Virtual machines (VMs) solve part of this challenge by providing secure, isolated, and scalable environments tailored to specific project needs. However, dedicating an entire physical GPU to a single VM is highly inefficient for mixed or lightweight workloads. 

This is where NVIDIA Multi-Instance GPU (MIG) technology becomes essential. With MIG, a single physical GPU is partitioned at the hardware level into multiple fully independent instances, each with guaranteed memory, cache, and compute cores. For a development team, this ensures predictable, uncompromising Quality of Service (QoS). This means that multiple developers can simultaneously train AI models, run simulations, or render graphics on the same physical server without competing for resources or interfering with one another’s workloads.

The NVIDIA RTX PRO 4500 Blackwell Server Edition GPU, featuring 32 GB of high-speed GDDR7 memory and support for up to two MIG instances, and the newly released NVIDIA vGPU 20 software deliver a substantial performance boost to accelerate diverse workloads across virtualized enterprise data centers. Together, they can power everything from everyday productivity to lightweight AI development simultaneously.

This post walks through how to deploy virtual machines (VMs) with NVIDIA vGPU and the NVIDIA RTX PRO 4500 Blackwell Server Edition. It covers setting up MIG with vGPU, sizing for enterprise workloads, performance comparison, and supplementary features.

Configuring RTX PRO 4500 Blackwell Server Edition for vGPU

The RTX PRO 4500 Blackwell Server Edition GPU with MIG technology revolutionizes virtualized workloads. The solution allows a single physical GPU to be securely partitioned into two independent GPU instances, each providing a dedicated 16 GB GPU framebuffer. This enables administrators to precisely allocate dedicated GPU resources, ensuring guaranteed quality of service for diverse virtual machines.

To follow along with the tutorial presented in this post, ensure your environment meets the core requirements outlined in Table 1. The tutorial uses VMware vSphere to set up MIG and vGPU.

Table 1. Prerequisites for deploying VMs with NVIDIA vGPU and NVIDIA RTX PRO 4500 Blackwell Server Edition GPU

Configure vSphere for vGPU and MIG profiles

This step walks through the process of configuring the hypervisor settings to enable vGPU and mixed profile features. Then a 4 GB profile is attached to an MIG slice of an RTX PRO 4500 Blackwell Server Edition GPU. This tutorial has three VMs on the host. A Windows 11 VM with a 4Q profile will be configured to demonstrate the process. Note that time-slicing within MIG will be supported on the VMware Cloud Foundation (VCF) soon. 

Start at the vSphere web interface. 

Select the host: Navigate to the ESXi host running the RTX PRO 4500 Blackwell Server Edition GPU Configure PCI devices: Navigate to Configure -> Hardware -> PCI Devices Find the GPU: Locate the RTX PRO 4500 Blackwell Server Edition card in the list of PCI devices Figure 1. Selecting the RTX PRO 4500 Blackwell in the VMware vSphere Graphics Devices list

Configure vGPU modes and mixed-size profiles

Next, adjust the Graphics Device setting for the RTX PRO 4500 Blackwell Server Edition GPU.

Device Type: Select Shared Direct  vGPU Mode: Select Mixed Size to enable different sizes of vGPU profiles to run on the same MIG instance Figure 2. Configuring the GPU for Shared Direct as the device type and Mixed Size as the vGPU mode

Attach a MIG-backed vGPU profile

After the initial VM creation, the next step is to attach a MIG-backed vGPU profile. To do this, first ensure the VM is powered off. Open the VM settings, select Add New Device, and then choose PCI Device.

Figure 3. Adding a new PCI device to the Windows 11 VM settings

For this scenario, assign a vGPU profile that is backed by a MIG slice. The RRO 4500 Blackwell Server Edition GPU supports a maximum of two MIG slices. Each of these slices can accommodate various vGPU profiles. 

As shown in Figure 4, select the nvidia_rtx_pro_4500_blackwell_dc-1-4q profile, which is a 4 GB profile. The presence of _dc-1- in the profile name signifies that this single vGPU is backed by one MIG slice on the RTX PRO 4500 Blackwell Server Edition GPU.

Figure 4. Selecting the specific 4 GB MIG-backed vGPU profile for the VM

With the VM settings now configured, the New PCI device shows the MIG-backed vGPU profile (Figure 5). You can now power on the VM.

Figure 5. Final hardware configuration showing the assigned 4Q vGPU profile

Provisioning VMs for various enterprise workloads

The NVIDIA RTX PRO 4500 Blackwell Server Edition GPU, when paired with vGPU software, establishes a flexible virtual platform that IT organizations can leverage to accelerate diverse, mainstream enterprise workloads. As a demonstration, our team configured three distinct VMs, each utilizing a different MIG-backed vGPU profile—3B, 4Q, and 16Q. These specific profiles are typically well-suited for applications such as knowledge work, entry-level workstations, and data science or machine learning tasks.

For further exploration, run nvidia-smi on the hypervisor host where the RTX PRO 4500 Blackwell GPU resides. The nvidia-smi utility is the NVIDIA System Management Interface, a command-line tool essential for monitoring and managing NVIDIA GPUs. Running it on the ESXi host allows administrators to see the real-time operational status, including details on GPU memory utilization, current temperature, and critically, the configuration and status of the MIG GPU Instances and the vGPU profiles running within them.

Figure 6. Monitoring GPU and MIG status through the nvidia-smi command-line utility

MIG GPU Instance 1 is running a 16Q vGPU profile. MIG GPU Instance 2 hosts both the 4Q vGPU profile, as shown in the nvidia-smi output for this example, and an illustrative 3B vGPU profile.

It will be necessary to install the appropriate Windows 11 vGPU driver. Be sure to disable any software-emulated graphics that would interfere with the NVIDIA guest driver operation. Connect to the VM using a remoting protocol such as Omnissa Horizon or RDP. Once connected, open the Windows Device Manager and check the Display Adapters. Verify that the VM is using the MIG-backed vGPU profile, as configured previously in this tutorial.

Figure 7. Verifying the RTX PRO 4500 Blackwell vGPU profile in Windows Device Manager

Another experiment involves running the dxdiag utility. You can access this tool through the Windows search bar or by typing dxdiag into the command prompt. The primary graphics device should be identified as the RTX PRO 4500 Blackwell DC-1-4Q profile.

Figure 8. Confirming the RTX PRO 4500 Blackwell vGPU is active with full hardware acceleration

vGPU solution architects at NVIDIA experimented with running two applications simultaneously on separate VMs, demonstrating the versatility of the RTX PRO 4500 Blackwell GPU with MIG-backed vGPU. One VM hosted a CUDA-based particle simulation running in a Linux Docker container, while the other provided a Windows 11 desktop for a knowledge worker. These two distinctly different workloads were executed simultaneously.

Figure 9. Simultaneous execution of a Linux-based CUDA particle simulation and a Windows 11 enterprise desktop

Accelerate workloads with NVIDIA RTX PRO 4500 Blackwell Server Edition

For organizations currently using NVIDIA Ampere (A-series) or NVIDIA Ada Lovelace (L-series) architectures, the move to NVIDIA RTX PRO 4500 Blackwell Server Edition is a transformative leap rather than a marginal gain, particularly for AI-augmented applications. Specifically, in a virtualized environment, the RTX PRO 4500 Blackwell Server Edition provides nearly 1.9x the acceleration for graphics workloads in a 4K setup compared to the NVIDIA L4.

Figure 10. The RTX PRO 4500 Blackwell Server Edition GPU doubles the performance of the L4 in SPECviewperf 15 at 4K throughput

Enterprise knowledge workers require a responsive and interactive desktop experience, even as organizations scale their infrastructure. The RTX PRO 4500 Blackwell Server Edition GPU provides a modern platform designed for these deployments. The Blackwell architecture introduces capabilities such as MIG, which spatially partitions the GPU to deliver predictable performance, improved resource utilization, and reliable quality of service for virtual desktops. Beyond raw compute and partitioning, RTX PRO 4500 Blackwell also delivers higher NVENC throughput, enabling more efficient remote display streaming.

With the launch of vGPU 20.0, RTX PRO 4500 Blackwell Server Edition is fully integrated with major virtualization platforms, providing a seamless, deployment-ready solution for enterprise IT environments. This update ensures that IT admins can leverage the RTX PRO 4500 Blackwell Server Edition architectural advancements within their data center deployments.

More enhancements in vGPU 20

In addition to supporting NVIDIA RTX PRO 4500 Blackwell Server Edition, vGPU 20 introduces the following enhancements designed to accelerate modern applications and ensure consistent performance within virtualized environments:

New NVIDIA AI Virtual Workstation (vWS) Toolkit: Deploys an NVIDIA AI Blueprint for Video Search and Summarization (VSS) in a vWS environment running on NVIDIA RTX PRO 6000 Blackwell Server Edition Fixed-share scheduling for heterogeneous vGPU: Ensures consistent scheduling duration and frequency for each vGPU instance irrespective of other instances on the physical GPU  vGPU support for VergeOS: Automates GPU driver management, vGPU profile assignment, and MIG configuration through its UI within a private cloud operating environment Wayland support: A display server protocol for Linux-based virtual machines Liquid-cooled GPU support for NVIDIA RTX PRO 6000 Blackwell Server Edition 

NVIDIA vGPU and NVIDIA Blackwell availability in the cloud

GPU virtualization offers a cost-effective way for enterprises to access necessary GPU resources through VMs from any location, utilizing only the capacity they require. For demanding AI and visual computing tasks that require high GPU memory, major cloud server providers now offer NVIDIA vGPU and NVIDIA Blackwell-powered instances:

Google Cloud: Announced the preview of fractional G4 VMs powered by NVIDIA RTX PRO 6000 Blackwell Server Edition. These vGPU-enabled configurations include 12 GB, 24 GB, 48 GB, and 96 GB profiles, supporting use cases from streaming services to high-fidelity 3D rendering and robotics sensor simulation. Microsoft Azure: The NCv6 series, powered by NVIDIA RTX PRO 6000, will offer NVIDIA vGPU-enabled fractional and full GPU options up to dual 96 GB.

Get started with NVIDIA Blackwell and NVIDIA vGPU

Harness the power of NVIDIA Blackwell virtualization to transform your data center for the AI-infused era of enterprise computing. By leveraging the NVIDIA RTX PRO 4500 Blackwell Server Edition and NVIDIA vGPU 20, your organization can use its existing, proven infrastructure to support an expanding variety of applications, as every application increasingly integrates AI capabilities.

Ready to see the difference? Request your free 90-day NVIDIA vGPU software trial.