NVIDIA GRID™ OVERVIEW
Imagine if responsive Windows and rich multimedia experiences were available via virtual desktop infrastructure, even those with intensive graphics needs. NVIDIA makes this possible with NVIDIA GRID, a virtualized GPU that delivers a faster, interactive graphical experience to a variety of users across all devices.
Why GPUs for Virtual Desktops and Applications?
Your smartphone has a GPU, your Tablet has a GPU, your PC has a GPU and your workstation has a high powered GPU. There's a reason for this. Users expect a certain level of visual experience that cannot be obtained from a CPU alone. GRID enables VDI for everyone.
What is a Virtual Desktop?
Virtual Desktop Infrastructure (VDI) is the practice of hosting a desktop OS within a virtual machine (VM) running on a hosted, centralized or remote server. It allows IT departments to offer a secure "access anywhere" desktop experience to end users.
What is a Virtual Application?
A virtualized application, more commonly referred to as a virtual user session, allows multiple users to run applications in a single OS (e.g., Windows Server).
Why NVIDIA?
Continuing its tradition of pioneering technology, NVIDIA's Kepler Architecture is the world's first virtualization-enabled GPU. NVIDIA GRID K1 & K2 boards bring this technology to the widest breadth of partner products and solutions.
Visual representation of a server rack connected to various client devices (desktop computer, tablet, smartphone), with text indicating 'SERVER POWERED BY NVIDIA GRID' and 'GPUS ACCELERATE VIRTUAL DESKTOPS AND APPLICATIONS'.
Who Needs NVIDIA GRID™?
Designer/Engineer
Traditionally an NVIDIA Quadro® user, this customer creates and works with complicated datasets using graphics-intensive applications (3D design, medical diagnostics, etc.).
Related industries: Oil & Gas, Manufacturing, Media & Entertainment, Medical Imaging, Public Sector
Application examples: Dassault Systèmes (CATIA, SolidWorks), Enovia, Siemens NX, Autodesk (3ds Max, Inventor, Maya, etc.)
*Plus all the same applications a Power User and Knowledge Worker can use.
Power User
A user of visual data (such as 3D images, and 2D graphs and line charts). This customer represents a new business opportunity. The customer may also be someone who tried Desktop and Application Virtualization without GPU acceleration and was not satisfied.
Related industries: Healthcare (nurses' stations, doctors' offices, doctors' tablets), Educational Institutions (target engineering and design schools), Government (simulation and training, Geospatial research), Manufacturing (Product Data Management, Product Lifecycle Management, Manufacturing Floor/Job Site, Support).
Application examples: SolidWorks View, PTC Creo View and View Express, Autodesk Inventor View, Siemens NX Viewer, Team Center Visualization. Manufacturing Product Lifecycle applications, Adobe Imaging applications, medical imaging applications, Microsoft Office PowerPoint Authoring, Video editing applications.
**Plus all the same applications a Knowledge Worker can use.
Knowledge Worker
This is typically an office worker who uses Office applications, email, video conferencing, and rich media internet applications.
Related industries: Financial Services (Retail, Commercial and Investment Banking, Insurance), Manufacturing, Life Sciences, Oil & Gas, Media & Entertainment, Telecommunications, Government, Education, Technology.
Application examples: Microsoft Office (PowerPoint graphics), Photoshop, etc.
Examples of Industries Where GRID Can Be Used
- Financial Institutions
- Government
- Healthcare Providers
- Media & Entertainment
- Manufacturers
- Educational Institutions
NVIDIA GRID for VDI and Application Virtualization
NVIDIA GRID Platforms for VDI
Server: Contains Application, GRID Enabled Virtual Desktop, NVIDIA Driver, Guest OS, Virtual Machine, GRID Enabled Hypervisor, and NVIDIA GRID GPU.
Client Devices: Various devices connecting to the virtualized environment.
Diagram showing a server hosting a Virtual Machine with a GRID Enabled Virtual Desktop, NVIDIA Driver, and Guest OS, connected to client devices. A GRID Enabled Hypervisor and NVIDIA GRID GPU are foundational elements.
NVIDIA GRID Platforms for Applications
Server: Contains multiple Applications (App) running in XenApp Sessions, managed by an NVIDIA Driver and Server OS within Virtual Machines, supported by a GRID Enabled Hypervisor and NVIDIA GRID K1 BOARD.
Client Devices: Various devices connecting to the virtualized environment.
Diagram showing a server hosting Virtual Machines running XenApp Sessions, managed by an NVIDIA Driver and Server OS, connected to client devices. A GRID Enabled Hypervisor and NVIDIA GRID K1 BOARD are key components.
Glossary of Terms
| Term | Description |
|---|---|
| Citrix Receiver | A lightweight software application that runs on the user's Windows, Mac, Linux, iOS, and Android devices and connects to a XenDesktop virtual machine or XenApp session in the data center. |
| Citrix XenApp | Citrix's virtualization product that provides remote session capabilities, delivering remote applications or virtual sessions. |
| Citrix XenDesktop | Citrix's desktop virtualization product that provides remote desktop capabilities. |
| Citrix XenServer | The commercial hypervisor from Citrix that allows multiple operating systems to run on a single server node. |
| Cloud Computing | The use of computing resources (hardware and software) that are delivered as a service over a network. Clouds can be public (delivered over the internet) or private (contained within a company's intranet). |
| Dedicated GPU | A case where the full GPU is used by a virtual machine, not sharing it with other VMs. |
| GPU Pass-Through | The remoting technology that connects a virtual machine (VM) to a dedicated CPU. The technology to do this from NVIDIA is known as NMOS (NVIDIA Multi-OS), from the capability to have multiple OS's running on a system, each with their own GPU. |
| GRID | NVIDIA technology that is a combination of both hardware and software to deliver the ultimate virtualized experience. |
| GRID K1 | NVIDIA's GRID K1 is a graphics card designed specifically for virtualization use cases. It carries four Kepler GPUs and a total of 16GB of video memory, making it ideal for multi-user configurations. |
| GRID K2 | NVIDIA's GRID K2 is a graphics card designed specifically for workstation virtualization. It offers two high-end Kepler GPUs, which can support a total of two high-end designers and engineers. The GRID K2 can also be used for power users and knowledge workers, especially where the IT department wants to deploy a single resource type across the entire enterprise. |
| Host Machine | A computer on which a hypervisor is running one or more virtual machines is a host machine. Each of those virtual machines is called a guest machine. The hypervisor presents to the guest operating systems a virtual operating platform and manages the execution of the guest operating systems. |
| Hypervisor | In computing, a hypervisor or virtual machine manager (VMM) is a piece of computer software, firmware or hardware that creates and runs virtual machines. |
| Kepler | The world's first cloud enabled GPU, and the latest architecture from NVIDIA. NVIDIA products built on the NVIDIA® Kepler™ architecture have names such as NVIDIA Quadro® K5000, NVIDIA Tesla® K20, and GRID K2 (K for Kepler). |
| Microsoft RemoteFX | Microsoft's product that enables multiple users to have access to the acceleration of a single hardware GPU. It is available as a feature of Microsoft Server 2008 R2 and Server 2012. |
| NVIDIA Windows Server Driver | The NVIDIA graphics driver that loads on a server OS. This driver works with RemoteFX to deliver graphics acceleration. |
| NVIDIA ESX Driver | This NVIDIA driver runs on the ESX (otherwise known as vSphere) hypervisor, which allows VMs running on top of ESX to get access to a shared NVIDIA GPU for hardware graphics acceleration when using vSGA (see definition below). |
| NVIDIA Driver | NVIDIA's certified driver that runs inside the virtual machine, giving graphics acceleration directly to the applications. |
| Remote Workstation | A workstation instance that is running from a data center, and delivered via the network to a client device. "Remote" can be on premise to the user's office, or can be off premise to a partner site, travel location, or home user. |
| Virtual Machine | The instance of an OS that runs on top of a hypervisor, using the abstracted hardware image presented by the hypervisor. |
| Virtualization | The practice of abstracting the virtual machine (VM) from the physical hardware that it runs on. In practice, virtualization is used primarily to run multiple VMs on a single piece of physical hardware. |
| Virtual Desktop Infrastructure (VDI) | The practice of hosting a desktop operating system within a virtual machine (VM) running on a hosted, centralized or remote server. |
| Hardware Virtualization | The creation of a virtual machine that acts like the real hardware running underneath the hypervisor, or a subset of that hardware. The software executed on these virtual machines recognizes the underlying physical hardware resources (i.e. the OS can load a native driver for that hardware and interact directly with it). |
| Virtual User Session | The practice of hosting applications in a single server OS and delivering those applications to multiple remote users. |
| Hardware Virtualized GPU | NVIDIA Kepler-based K1 or K2 board that allows multiple users to share a single GPU and allowing each user direct access to the GPU hardware. This improves user density while providing true performance and compatibility. |
| Software Virtualization | Software virtualization acts as an interface between the hardware and the VM, creating the ability to adapt to different underlying hardware configurations. In this case, the software acts like the hardware resource, passing commands to the hypervisor which can execute them either on real hardware or emulated hardware. |
| NVIDIA GRID Virtual GPU (NVIDIA GRID vGPU) | Key technology used to enable the hardware virtualization of the GPU which allows multiple virtual machines to interact directly with a GPU. The GRID Virtual GPU manages GPU resources that enable multiple users to share common hardware while improving user density, and creates a true PC experience in the cloud. |
| VMware ESX | The commercial hypervisor from VMware that allows multiple operating systems to run on a single server node. |
| VMware View | VMware's desktop virtualization product that provides remote desktop capabilities. |
| VMware vDGA | Virtual Dedicated Graphics Acceleration. VMware's productization of GPU pass-through in ESX. Each GPU in the system is dedicated to a single VM, and the NVIDIA native graphics driver is loaded in each of those VM's operating systems. |
| VMware vSGA | Virtual Shared Graphics Acceleration. The software virtualization of a GPU in ESX, allowing multiple users to access a single GPU for graphics acceleration. In this case, the NVIDIA ESX driver is loaded in the hypervisor. |
Diagram illustrating GPU pass-through technology, showing multiple operating systems on workstations each with a dedicated GPU.
Diagram highlighting NVIDIA's Kepler architecture, noting it as the world's first cloud-enabled GPU with billions of transistors.
Diagram indicating NVIDIA's long-standing support for Windows Server graphics drivers, dating back to 2003.
NVIDIA GRID™ Board Specifications
NVIDIA GRID boards are designed for higher GPU densities, with 2 or 4 GPUs per card. For example, with a GRID K1 card, there are 4 GPUs per card. A server with two x16 PCIe slots can hold 2 of these cards, for a total of 8 fully accelerated VMs.
| Feature | GRID K1 | GRID K2 |
|---|---|---|
| Number of GPUs | 4 entry Kepler GPUs | 2 high-end Kepler GPUs |
| Total NVIDIA CUDA cores | 768 cores | 3072 cores |
| Total GPU memory size | 16 GB DDR3 (4 GB per GPU) | 8 GB GDDR5 (4 GB per GPU) |
| Max power | 130 W | 225 W |
| Board length | 10.5" | |
| Board height | 4.4" | |
| Board width | Dual slot | |
| Display IO | None | |
| Aux power | 6-pin connector | 8-pin connector |
| PCIe | x16 | |
| PCIe generation | Gen3 (Gen2 compatible) | |
| Cooling solution | Requires GRID certified server, passive cooling | |
| Graphics APIs | Open GL 4.3, DirectX 9, 10, 11, CUDA |
