NVIDIA is the reference standard for Enterprise PC’s and workstations with 90% market share for workstations and 60% for PC’s. Graphically demanding workflows are built on an NVIDIA graphics platform and all applications run with NVIDIA trusted quality and performance. Workstation applications are optimized and certified for Enterprise IT to reliably deploy across the organization for an uncompromized user experience for the most demanding applications. For PC and workstation virtualization, GRID enables “as local” experience for graphically demanding workflows by fully exposing the GPU within leading hypervisors from VMWare and Citrix
In the Datacenter, NVIDIA Tesla GPU’s are deployed for high performance computation for analysis and big data.
All of these changes have brought us to the future of end user computing. A world where IT can support their users anywhere, on any device with any applications. Removing barriers for productivity so they can accelerate the business.
This is where NVIDIA, the leader in graphics acceleration, stepped in to help. NVIDIA GRID technology allows IT to virtualize the physical GPU sitting in a server and share it with multiple VDI instances. This means that IT can deliver a true PC experience to any remote device from the datacenter. By providing a way to bring graphics acceleration to virtualization, NVIDIA GRID allows you to unlock all of the promises of productivity, mobility, security and flexibility for every one of your users.
With NVIDIA GRID you can safely house ALL your current into the datacenter so that they can be delivered out to any device, be it a thin client, chromebook, iPad or BYO Device. From an end user perspective this means they can be more productive working with the devices and in the locations that best suit them. For IT they are able to manage everything in a centrally in the datacenter which vastly simplifies their life.
So let’s see what we mean by when we talk about the challenges of traditional VDI. As you can see the image on the left is an example of a VDI deployment that only can show a basic model which has been unacceptable to users and unable to manage complex objects. Because this machine is leveraging GPU graphics emulation ie non native hardware support not all application features can even be supported.
NVIDIA GRID exposes a full PC experience for the industry leading virtualization environments. Any virtualized PC or workstation application now has 100% of the functionality and performance of a local machine. The virtualized image on the right shows an image displaying full performance and quality allowing users to visualize all the components of the model. NVIDIA GRID enables the performance and visual quality you would expect on a local device.
How does this solution actually work? We talked about virtualizing the GPU and being able to deliver that application compatibility to any device with that native experience that users are used to. We do this using our vGPU technology to divide up that physical GPU and then deliver that straight through to the actual virtual machine. The graphics commands of each virtual machine are passed directly to the GPU, providing them with the graphics resources they need to launch and use their applications in whatever client device they choose to use.
There is no other product or solution on the market today that can do this. NVIDIA GRID is the only product that can deliver accelerated virtual desktops so that users can be freed up to work no matter what device they’re using or no matter where they are.
At the same performance level as GRID K2, you can have twice as many users with the Tesla M60 and GRID 2.0.
Note that on Viewperf, the average performance is slightly higher with 32 VMs on M60 as compared with just 16 VMs on K2.
While Viewperf is an artificial benchmark designed to test mostly graphics, ESRI’s ArcGIS Pro is a professional 3D application. Using their simulated user environment, which tests the entire platform, we are able to gauge the scalability of the entire system. With the new GRID 2.0, we again achieve twice the number of users (32 vs 16) with only a slight degradation in performance (note that the second chart is measured in time, so lower is better).
If we keep the number of users consistent between GRID 1.x and 2.0, we can see that the raw performance of the M60 over K2 is 2X. In this case, we measured single user Viewperf 12 raw performance comparing the two platforms. Here we varied the vGPU profiles to see the impact across the multiple options. You can see that in every case, performance is 2X or nearly 2X. In the 8GB framebuffer case, you can’t even run on a K2 because of the 4GB max memory on that card.
If raw performance is important, and increased scalability is not a requirement, then Tesla M60 with GRID 2.0 will deliver up to 2X the result.
Moving Quadro to the datacenter As a result you need 5x 9’s support
Here are the features of the 3 GRID software editions, listed down the vertical axis, GRID Virtual PC, GRID Virtual Workstation, and GRID Virtual Workstation Extended. We have a number of displays and maximum resolution that we support for each one of these. Your business users tend to use a standard single or dual virtual display up to 2560x1600 resolution. In the virtual workstation space, we start to see up to 4 virtual displays being used, and we even support 4k in the virtual workstation extended edition. All 3 editions support Windows Guest OS, and add Linux guest OS support in the GRID Virtual Workstation and Virtual workstation extended editions. Our traditional Quadro software features are supported in the two Workstation class products as well. We add support for CUDA and OpenCL in the GRID Virtual Workstation Extended edition in our 8GB vGPU profile, as well as offering pass-through. You can see the vGPU profiles that are supported in each edition on the right side of the table. So, up to 2GB profiles in the Virtual PC edition as well as the Virtual Workstation edition, and then the extended goes up to 8GB profiles.
NVENC K2 x6 (1080p30)
We recently introduced our new Maxwell hardware for accelerated virtual desktops at VMworld, and those two pieces of hardware, the Tesla M60 for PCI-express-based rack and tower servers, and the Tesla M6 is our blade server architecture. The difference between the two is that the M60 is a dual-GPU high-end Maxwell, and the M6 is a single high-end Maxwell. Compared to K2, that’s a dual high-end Kepler, so you can see that M60 is really a direct replacement for K2, and the M6 is essentially half of the M60 on a single board. The performance numbers are doubling this generation (2x K2) for M60. From a scalability standpoint, you can see that K2 supported up to 16 users, and M60 doubles that up to 32 users. For example, where you could support 8 users on a K2, you can now support 16 users on an M60 at the same performance level.
With software, there will be updates, and with our subscription, we’ll offer ongoing access to updates for our GRID software so you can take advantage of the latest features and improvements, even the latest hardware enablement that we make in our software. Since it’s a perpetual license, you can get updates to support the latest hardware as long as you’re current with your subscription. And then of course, maintenance is always part of software. You have to do long-term resolution of defects and security fixes…anything that comes out will be enabled as part of this subscription.
There are 2 versions of SUMS, a basic version and a Production version. The differences are generally the level of service that we offer.
With both of these, you’ll get the latest maintenance releases and latest version upgrades as they become available. The difference is the long-term branch maintenance. With a basic subscription, we’ll continue branch maintenance for up to 1 year from general availability of that branch release. With production, we’ll support that up to 3 years from general availability so we can add fixes and maintenance and features to that. Response time difference between basic and production is 1 business day vs 4 hours when we’ll get back to someone. And then access is a little bit different. If you want phone access, that’s only available in the production version, as well as the 7 by 24 capability versus standard business hours in the basic edition. So those are the two editions and the differences for SUMS.
Users whose support contract is current see latest drivers and updates Users without a current support contract only have access to drivers/updates that were released during the lifetime of their prior support contract Exception: releases that incorporate critical security fixes may be made available to users who are not current on support (i.e. they get a one-time, free update in order to pick up security fixes)
Access to the latest improvements delivered as Updates and Upgrades Long-term Support and Maintenance for NVIDIA GRID software Direct communication with NVIDIA technical experts Searchable knowledge base of specific documentation, application notes and articles Rapid response & timely issue resolution Active issue prioritization
The key value of NVIDIA GRID lies in bringing the benefits of virtualization to all of your users. While many organizations have started seeing the benefit of virtual PCs with their lowest level users, the value greatly increases as you add more use cases. To look at how these benefits are realized I’d like to walk you through a few customer examples:
This technology, which has been around since 2013, has seen adoption across numerous industries. These are just a handful of the organizations that are looking at how NVIDIA GRID can help them deliver accelerated virtual desktops across their organization.
End user productivity: Metro Health - Regional healthcare provider in Grand Rapids Michigan who began deploying VDI 8 years but were not satisfied with results. Currently testing GRID K1 boards and vGPU with 95 users, with plans to eventually deploy to all 4,500 users. Applications include Epic, clinical applications, office and web browsing. The current VDI implementation has been proven through internal and external analysis to be delivering a time savings of 30 minutes per day to each doctor and 50 minutes per day to nurses and other professionals at Metro Health through fast roaming access alone.
IT management: NC State University’s more than 33,000 students and nearly 8,000 faculty and staff, NCSU Library IT set up the Horizon View infrastructure with two graphics processing unit (GPU)-enabled clusters that include Dell servers and NVIDIA GRID graphics cards to run Solidworks and AutoCAD applications. Prior to virtualization, the library had standard Windows desktops. All applications were pushed out directly to physical desktops by campus IT staff. Now provisioning a new desktop takes very little time, and typically 10 or so standby virtual machines are available—so when someone logs out, a new one is created with several remaining at any given time.
Data security: SSOE Group – large engineering, procurement and construction management firm with 1,200 employees; 29 offices in seven countries. Supporting collaboration among employees, contractors, and third parties globally— working on centralized design and 3D models—created a real technology challenge for SSOE. End users required reliable and fast access to critical design and engineering applications such as Autodesk Revit, Autodesk AutoCAD, the Autodesk design suite, Bentley Design Suite, and SolidWorks 3D CAD design software. The firm needed to support big, complex models that were part of projects that could have 70 or 80 models connected together, and the end-user model viewing and manipulating experience had to have responsive performance. SSOE can leverage all of the architectural, engineering, and construction expertise in the world without compromising the security of SSOE or client intellectual property.
GRID 2.0 ソフトウェアエディション
最⼤画⾯数 2 4
画⾯あたりの最⼤解像度 2560 x 1600 2560 x 1600
Windows ゲストOS ü ü
Linux ゲストOS ü
NVIDIA® Quadro® ソフトウェア機能 ü
対応の vGPU プロファイル (フレームバッファとGPUあたりの最⼤ユーザー数)
512 MB (GPUあたり最⼤16ユーザー) 2
1 GB (GPUあたり最⼤8ユーザー)2
2 GB (GPUあたり最⼤4ユーザー) ü ü