RemoteFX Relies on Servers with GPUs

By Andrew Garcia  |  Posted 2010-06-14 Print this article Print


RemoteFX, meanwhile, promises users a much richer client experience over VDI than was previously possible. RemoteFX, a byproduct of Microsoft's 2008 acquisition of Calista Technologies, delivers three-dimensional graphics rendering for Windows 7 virtual desktops, allowing users to take advantage of Aero desktop features, watch high-definition video and use 3D-intensive applications. RemoteFX also adds support for a wider range of client-connected USB devices, such as multifunction printers/scanners and unified communications devices such as headsets or Webcams.

The new graphics capabilities delivered via RemoteFX will likely not be realized by customers in the near term, as the feature requires a number of hardware and software enhancements on both the virtual server and client endpoints.

First of all, RemoteFX requires that host servers have a GPU installed. Graphics rendering for all virtual clients happens on the server host, with the server GPU being exposed to the guest operating system via a virtual GPU driver. RemoteFX time-slices the server's GPU to share across multiple Windows 7 remote or virtual sessions. Rendered data is sampled, then tightly compressed using a proprietary codec and delivered as a bit map over the network back to the virtual client via RDP. Compression at the server may be done in software (with work done by both GPU and CPU) or in hardware, if the hardware has an accelerating ASIC (application-specific integrated circuit) on board.

Microsoft claims RemoteFX uses intelligent screen capture to sample only detected changes, rather than the whole image over and over, but bandwidth usage is still enough to make the feature usable only over LANs. RemoteFX will send more frames on really fast networks, though. Administrators will be given centralized control over the maximum resolution and number of monitors supported on virtual clients, allowing some predictability in network capacity planning when using RemoteFX. Performance planning and deployment guides should be available by the SP1 RTM timeframe.

During a breakout session about RemoteFX, Justin Graham, senior product manager in the Windows Server Group, acknowledged that no one currently has GPUs in their servers (in a show of hands of the audience, only one member reported using GPUs in a server.) Graham said Microsoft wanted to deliver RemoteFX as soon as possible, in order to allow its customers time to adjust to, plan for and adopt the new hardware requirements for RemoteFX. And for its hardware partners to deliver those technologies in the first place so customers can buy them.

According to Microsoft's recommendations, each server in a RemoteFX-enabled VDI pool needs its own GPU (or an external GPU appliance can be used) to ensure RemoteFX operation as VMs migrate between hosts. Identical GPUs are needed if VM migration is a requirement.

RemoteFX is intended for new VDI deployments due to the increased hardware demands, which will require new servers in most cases. Customers should also account for the additional heat and power complexities that may arise with the use of server GPUs. Fortunately, Microsoft claims, RemoteFX is GPU-hardware-agnostic, so ATI, Nvidia and Intel GPUs could be equally effective.

Server-side graphics processing could be beneficial for low-horsepower clients such as netbooks or thin clients that lack higher-end graphics cards. Of course, the endpoints will need to support RemoteFX as well. For Windows 7-based endpoints, SP1 delivers a new Remote Desktop client (Version 7.1) that adds this support, and there should be updated clients available for Vista and Windows XP SP3 as well. Thin clients will likely need to be swapped out for newer models that support the RemoteFX codec on an ASIC, however.

Microsoft System Center users should be aware than these new features will not immediately be configurable through Virtual Machine Manager, although Microsoft officials said they expect support to be added soon after launch.

For non-Hyper-V or VDI customers, SP1 adds a few other new features. For instance, improvements to Active Directory increase the maximum number of connections, helping performance particularly in networks with high latency. Also with SP1, DirectAccess adds support for 6to4 and ISATAP addresses when used with Microsoft's Network Load Balancing.

Interestingly, I learned from Bill Laing, corporate vice president of Microsoft's Windows Server and Solutions division, that DirectAccess wasn't immediately appealing to the anticipated audience. Laing said, "We thought DirectAccess was pretty much an enterprise feature, but it's actually been small and medium businesses that love it most."

When I asked whether these customers were tightening security and access controls via NAP (Network Access Protection), he responded, "No, they are keeping it pretty simple."

For more information about the changes coming in SP1 and deeper dives into the underlying technologies, check out these TechEd sessions (VIR304, VIR305, WSV307, WSV309, WSV13INT).

Andrew cut his teeth as a systems administrator at the University of California, learning the ins and outs of server migration, Windows desktop management, Unix and Novell administration. After a tour of duty as a team leader for PC Magazine's Labs, Andrew turned to system integration - providing network, server, and desktop consulting services for small businesses throughout the Bay Area. With eWEEK Labs since 2003, Andrew concentrates on wireless networking technologies while moonlighting with Microsoft Windows, mobile devices and management, and unified communications. He produces product reviews, technology analysis and opinion pieces for, eWEEK magazine, and the Labs' Release Notes blog. Follow Andrew on Twitter at andrewrgarcia, or reach him by email at

Submit a Comment

Loading Comments...
Manage your Newsletters: Login   Register My Newsletters

Rocket Fuel