Nvidia GRID K1 User manual

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. The users expect a certain level of visual
experience that you can’t get 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 (ie: Windows Server).
WHY NVIDIA?
Continuing our 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.
GPUs ACCELERATE
VIRTUAL DESKTOPS
AND APPLICATIONS
SERVER POWERED
BY NVIDIA GRID
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.
NVIDIA GRID™
OVERVIEW
GRID OVERVIEW

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. 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
EDUCATIONAL INSTITUTIONSMANUFACTURERS
MEDIA & ENTERTAINMENTHEALTHCARE PROVIDERS
GOVERNMENTFINANCIAL INSTITUTIONS
+ +
+

CLIENT DEVICES CLIENT DEVICES
GRID Enabled Hypervisor
NVIDIA GRID GPU
Guest OS
Virtual Machine
Application
GRID Enabled Virtual
Desktop
NVIDIA Driver
SERVER
Server OS
Virtual Machines
XenApp
Session
App
XenApp
Session
App
XenApp
Session
App
XenApp
Session
NVIDIA Driver
SERVER
App
GRID Enabled Hypervisor
NVIDIA GRID K1 BOARD
NVIDIA GRID FOR VDI
AND APPLICATION
VIRTUALIZATION
NVIDIA GRID PLATFORMS FOR VDI NVIDIA GRID PLATFORMS FOR APPLICATIONS
ARCHITECTURAL OVERVIEW

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.
GLOSSARY OF TERMS
NVIDIA has been
supporting, shipping
and certifying
a Windows Server
graphics driver
since 2003.
Kepler, the world’s
first cloud enabled
GPU, has up to 7 billion
transistors.
GPU pass-through
was first introduced in
2009 to deliver multiple
hardware accelerated
operating systems on
workstations.
GLOSSARY

TERM DESCRIPTION
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.
GLOSSARY OF TERMS CONTINUED
NVIDIA GRID enables
a true PC experience
for the hundreds of
millions of power users
who increasingly want
to bring their own
devices to work.
The GRID Virtual
GPU technology is
integrated into a
commercially available
system level hypervisor,
like Citrix XenServer.

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
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.
NVIDIA GRID™BOARD
SPECIFICATIONS
SPECIFICATIONS