Microsoft’s secret plans for RemoteFX: Azure-based desktops, apps, and Xbox games from the cloud?

By now everyone should be familiar with Microsoft's upcoming "RemoteFX" extension to RDP which will be available as part of Windows 7 and Windows Server 2008 Service Packs 1. RemoteFX promises to offer a near perfect remote display experience, including multiple displays, 3D, multimedia, and Aero glass.

By now everyone should be familiar with Microsoft's upcoming "RemoteFX" extension to RDP which will be available as part of Windows 7 and Windows Server 2008 Service Packs 1. RemoteFX promises to offer a near perfect remote display experience, including multiple displays, 3D, multimedia, and Aero glass.

The downside of RemoteFX is that it will require some serious host-side GPU processing, and that in the first release it will only be targeted for LAN scenarios.

It's no secret, though, that Microsoft plans for RemoteFX to be as ubiquitous as possible. In fact they'd like it to become as popular as H.264, with RemoteFX becoming the de facto standard for live-generated interactive content (screens, apps, games, etc.), while H.264 will be used for pre-rendered video content (movies, tv shows, youtube, etc.).

One of the qualities of RemoteFX is that while the remote host encoding requires quite a bit of horsepower, the client-side decoding is relatively simple--something that can be done via a very low-cost chip or even by adding a couple hundred thousand logic gates to existing system-on-chip designs.

And to that end, Microsoft has already announced deals with LG where LG is building displays that have RemoteFX decoding capabilities built right in. So when you buy your fancy new LG 42" flat screen TV, the Ethernet port on the back will allow it to connect to a network to essentially become a huge RemoteFX thin client.

Of course this shouldn't be surprising. Teradici has already done a similar deal with Samsung who now has a line of displays with PC-over-IP decoding chips built-in. And just about every TV you buy nowadays has advanced capabilities which let it play video-on-demand from various websites, so the idea that TV makers would add RemoteFX or PCoIP capabilities as standard offerings in the next few years is not too far fetched.

So what's this have to do with Azure?

Ok, so far, so good. But what's the point? Well so far I don't feel like we've really gotten a good answer from Microsoft as to why they bought Calista and developed RemoteFX.

Does Microsoft really care about enabling a great remoting experience for Windows? I mean they've been fine to let companies like Citrix and Quest extend and enhance RDP for the past decade--why the sudden urge now for Microsoft to have to do this themselves?

From a public standpoint, Microsoft has stated the goal of RemoteFX is to "push Hyper-V sockets," which means "since Hyper-V is required for RemoteFX, we want to make this awesome RemoteFX thing so that everyone will want to use Hyper-V." And I gotta say, for the past year or so, I believed that. I believed the reason they created RemoteFX was to push Hyper-V seats.

But recently it hit me. "Wait.. What?!? Does that actually make sense? Does Microsoft really want to embed RemoteFX capabilities into endpoints and clients and TVs around the world just to push Hyper-V?"

Last weekend I visited friend-of-the-site Benny Tritsch in his home outside of Frankfurt, and the topic of RemoteFX came up. "Come on..." Benny said, "RemoteFX is not about Hyper-V, it's about Azure!"

Of course! Benny mentioned this just a few days after Microsoft announced their Azure-based infrastructure as a service (IaaS), which is their Amazon EC2-like offering where you can pay a few cents an hour to run Windows Server instances in the Azure cloud. (Gabe wrote about this last week.)

The more Benny and I discussed this, the more I felt he was right. The current "v1" of the Azure IaaS offering doesn't offer GPU support in the VMs, and thus doesn't offer RemoteFX support, but that's ok because the v1 of RemoteFX is not going to be aimed for the WAN anyway. But think about Microsoft's direction. Let's assume that in a few years the server vendors and GPU vendors actually have datacenter-tuned GPU offerings. And let's assume that Azure offers more control over individual VMs (and even support for Win7 VMs). And let's assume RemoteFX v2 works better on the WAN. (Well, and let's assume that every home user has multi-mbps bandwidth and is no more than 30-40ms from an Azure datacenter.

If those assumptions come true, then you have a pretty compelling framework for desktops and applications on-demand from Microsoft via Azure. (And hey, guess what! When this is all real we also have RemoteFX decode capabilities built-in to lots of different TVs, and the stand-alone RemoteFX thin clients are available at Best Buy for $99.)

How far away is that future? 2012? 2014? (Maybe we don't need to build the desktop on demand like I thought in 2015. Maybe Chetan is right?)

And by the way, don't think this stops with the Windows desktop and apps. Don't forget about Xbox from Azure. You'll just plug the Xbox controllers right into your TV. Or your $99 thin client. Games on demand. Apps on demand. You whole life in Azure. So let Google and VMware chase these new-fangled built-from-scratch Java/HTML5/whatever apps in the cloud. Microsoft will give you full rich Windows apps and games from the cloud, thanks to Azure and a little company called Calista that they bought almost three years ago.

Join the conversation


Send me notifications when other members comment.

Please create a username to comment.

This is a great article that will certainly stand the test of time. I couldn't agree more!

For the last several years, server-based computing has had one major hindrance: bandwidth.  It's the cost of bandwidth that has crippled the user experience.  Granted, there are other desirable management features, like layering, in search of a permanent solution, but it's chiefly the cost of bandwidth that has stymied the proliferation of hosted desktops.

Moore's Law states that the number of transistors per integrated circuit has roughly doubled every couple of years.  This means that every couple of years (12-18 months), a person can buy twice the computing power for the same amount of money.  However, this trend is expected to slow down beyond 2015, and possibly hit a brick wall by 2019.

Likewise, similar laws govern the economics of bandwidth and network capacity.  In this case, Butter's Law of Photonics states that the data capacity of optical fiber roughly doubles every 9 months, equating to 50% cost reduction per transmitted bit.  And that leads us to Nielsen's Law which states that the bandwidth available to users increases at a 50% rate per annum.

This is amazing stuff! Imagine if bandwidth costs could drop as precipitously as the costs of computing cycles.  It's no secret that the computing power found in an iPhone is greater than that of the supercomputer used in the Apollo 11 mission.  Now imagine bandwidth economics following the footsteps of computing economics.

We're clearly ushering in the age of Cloud Computing.  It's very possible, if not highly likely, that this could spur a major global economic revival.


If you look at what the likes of OnLive are doing hosting games in the (there, I said it!)this could easily become a reality over the next few years.

If MS can offer an XBox-compatible platform for developers via Azure, and all the consumer requires is a bluetooth controller talking to a RemoteFX-capable TV and a broadband connection; take-up could be massive.

No more expensive loss-making consoles to develop, distribute and maintain.

Your hardware resources (being in the data centre) can scale based on the type of game being played (dynamic CPU and memory allocation - we can do this today thanks to mass adoption of server virtualisation) and the current number of players. Small games will only get a tiny allocation, whereas complex games could get multiple cores, GPUs and large amounts of RAM. A "dynamic XBox" if you like.

Imagine the possibilities for developers...and the kind of games that could be developed when you arent constrained by the physical box sitting under your TV.

Exciting times indeed.


Delivering games remotely is a complex beast, As mentioned; Onlive is doing it and others are following suit.

Seamless integration with endpoints is paramount.

Lots of factors and variables here but I believe you're thinking in the right direction.


I mentioned the XBox in the cloud many moons ago somewhere on and no one bit.....


The focus of this discussion somehow turned to gaming. Why don't we just focus on corporate desktops and applications for now?


edgeseeker.... because IT is boring and we need something to keep us sane



Back to Edgeseeker's comment on the need for lower cost bandwidth: Agreed but we're hearing that there is more to the matter than cost per Mbps.  Accepted also that ubiquitous deployment of cloud resources for virtualized enterprise systems and for consumers - video streaming, gaming, and the like - has to mean reliance on unmanaged network services (e.g., the WAN; the public Internet) and cheap bandwidth, right?  If so, when Edgeseeker's desired economics kick in (and they will, for certain) we will all fill our boots with that cheap and cheerful bandwidth but it is the quality of that bandwidth has always been and will continue to be a real challenge to cloud providers and users.  E.g., we know how TCP clamps down on bandwidth in reaction to poor network quality - in reaction to loss and latency - and we know what TCP clamping does to network application performance.

What's all this have to do with RemoteFX and Azure?  Aside from the academic connection, we have some interesting applied connections.  There are a couple of reports (and illustrative videos) on our recent comprehensive testing of the performance of RemoteFX and RDP 7.0 under a variety of network quality scenarios and with a variety of applications.

We would to invite your constructive criticism of our findings.  If you're game, please download the pdf portfolio at

So, RemoteFX There is an interesting test report


Why can't I still get HDX connect to a desktop and be done..... What's the value of the broker? Why are the functions coupled together?


Great article and I completely agree this is where it is heading but not everyone will have the bandwidth for this type of thing for a while yet.

In the mean time I can think of some very good uses for this in the home. I hope this is brought back as a feature in Windows 8 to match the server. I don't believe Remote FX actually requires the Hyper V as I believe you can run it on a non hyper V terminal services server (I may be wrong).

Anyway for the home I can see two great routes for Microsoft to take. Firstly Media Center Extenders didn't take off because they couldn't do everything a direct PC plugged into the TV could do. If Microsoft partners offer cheap thin clients with remote fx or tv's with it built in they can solve this issue and compete with things like Apple TV. This would need to use the second user profile like extenders currently do so anyone on the PC can continue working.

Secondly for Microsoft's tablet plans I could see them having RemoteFx built in to connect to your home machine not only could you stream content you could also potentially play your PC games.