Azure Brings True Virtual Graphics to the Cloud with NVv4
Do you know what's so great about bacon and eggs, burgers and fries, and peanut butter and jelly? They're all great together!
So are Nutanix Frame and the new Microsoft NVv4 AMD-powered virtual machines. In this post you will learn the technology details about the NVv4 - the first Azure cloud based instance type to use virtualized AMD GPUs. I'll also share insights into the benefits of using GPUs with Desktop-as-a-Service. You'll also get a preview of Frame and the Azure NVv4 in action.
Frame and Microsoft Azure Frame is a secure cloud platform designed to deliver applications and desktops to business users on any device. Clients, agents, receivers or plugins aren’t required on the endpoint device; all you need is an HTML5 browser. Uniquely, Frame is also not software. The service is cloud hosted, controlled and managed by Nutanix. Nutanix is solely responsible for operating, maintaining and developing the solution as a Desktop-as-a-Service (DaaS) platform. The flexibility of Nutanix Frame allows you to select your preferred infrastructure for running your applications and desktops (Azure, AWS, GCP or hybrid environments powered by Nutanix AHV). Microsoft Ventures (now M12) was one of the core investors in Frame prior to the acquisition of Frame by Nutanix and is a key partnership for our DaaS play.
Since early March 2020 Microsoft has made Azure NVv4 Virtual Machines generally available. This new Microsoft Azure series machine is powered by AMD EPYC ‘Rome’ CPUs and AMD Radeon Instinct MI25 GPUs. The NVv4 machines are the first AMD CPU and GPU instance types available in public clouds, which provides more choice for customers and partners. While new public cloud vendors are continuing to enter this space, this competition unlocks new customer use-cases and keeps cloud vendors sharp, engaged and innovative.
The NVv4 is able to partition the AMD GPU through the use of hardware based GPU virtualization using SR-IOV starting at 1/8th of a GPU with 2 GiB frame buffer to a full GPU with 16 GiB frame buffer. One of the core benefits of GPU partitioning is the ability to give the virtual machine a slice of the GPU capacity or even the full GPU capacity depending on the customer use case. NVv4 Virtual Machines support up to 32 vCPUs, 112GB of RAM, and 16 GB of GPU memory. Another benefit of the AMD GPU is there are no additional license costs and the instance has a competitive price point. More insights about NVv4 pricing are available here.
|Size||vCPU||Memory (GiB)||Temp storage (SSD-GiB)||GPU partition||GPU memory (GiB)||Max datadisks||Max NICs|
The new Azure NVv4 series will give Nutanix Frame customers a wider range of GPU options for their virtual desktop and application streaming needs. This unlocks a broader range of use cases with a very compelling price/performance ratio supporting not only high-end cloud workstations but also task and knowledge workers who can benefit using the NV4v4 or NV8v4 options.
Why are GPUs important for our virtual applications and desktops?
More and more applications we use everyday such as Office365, Internet browsers, and the Windows 10 OS benefits when a GPU is available. It isn’t only the workstations that benefit, but also 3D Graphics, CAD/CAM and HPC which also require and benefit from using a GPU. Office productivity and knowledge workers also benefit when a GPU is available - the user experience is just better.
Why isn’t everyone using GPUs or vGPUs in virtual applications and desktops?
The biggest disadvantage of adding a GPU with virtual applications and desktops is cost. That’s why it is great that AMD is entering the public cloud space with NVv4 series on Azure and delivering a cost effective machine with a GPU. By enabling virtualized GPUs in the cloud with Azure NVv4 and Nutanix Frame, we are able to deliver a whole new level of value that unlocks a much broader set of use cases.
For an even bigger impact and GPU adoption, we will see AMD entering the on-premises space and going mainstream in the near future. The biggest competition of NVIDIA - AMD or Intel GPUs is the CPU as it drives more innovation and lowers the price for NVIDIA.
Frame and Azure NVv4 - Preview
The Frame Remoting Protocol (FRP) ensures that the virtual applications and desktop presented in the HTML5 interface perform well and can easily handle graphic-intensive applications. FRP works best in Local Area Network (LAN) in mobile and in Wide Area Network (WAN) scenarios. Based on available bandwidth and network conditions such as latency and packet loss, FRP automatically adapts, delivering the best experience possible. When you want to render and display resource intensive workstation applications from e.g. Autodesk, Adobe, and ESRI, the experience is great delivering up to 60 frames per second even in a multi-monitor configuration.
In the preview below you will see Frame with the Azure NVv4 machine in action, also take a close look at the embedded video which gives good insights into user experience in a WAN scenario.
The video provides a preview of Nutanix Frame and Microsoft Azure NVv4 working together. The virtual desktop is running in Azure South Central US ~8000km or ~5000miles away from the physical endpoint. In this configuration the NV8v4 is used. The end user device is using a HTML5 browser, no client - agent - receiver or plug-in is installed and delivering great user experience over a WAN with 120ms latency. Check out the video here:
Try Xi Frame for Yourself
While Frame and Azure NVv4 is an (internal) preview I can imagine you want to see Frame in action yourself. Click the link to test out Xi Frame for free. You will get a great overview of both the user experience and admin experience. If you want to evaluate Frame and start a 30-day trial check out this page for more information. We’ll keep you posted on the final release, so stay tuned!