Citrix Virtual Design Studio enters beta. Requires OpenGL and x64. Wait... what?

Michael Chang, a marketing manager at Citrix responsible for this beta program, just emailed me a PowerPoint presentation about this product.

[UPDATE] Michael Chang, a marketing manager at Citrix responsible for this beta program, just emailed me a PowerPoint presentation about this product. I've attached it above.

Citrix Virtual Design Studio. Is this the product that was announced at Summit a few months ago that lets you design your Citrix environments virtually? Nope. That's Citrix Workflow Studio. (Hey, the Citrix product naming & marketing department is two-for-two in my book now!)

Citrix Virtual Design Studio is the final name for what was codenamed "Project Pictor," a set of extensions for Presentation Server that let it run OpenGL-based CAD applications via ICA. So "Virtual Design Studio" is for Computer Aided Design, virtually (since ICA is application virtualization now). Get it? (I don't know where the "studio" part came from either.)

Right now Virtual Design Studio only works with Dassault CATIA v5 running on the x64 edition of Windows Server 2003. From what I understand this is a full build of Presentation Server 4.5 x64 edition. (Well, the Citrix person I talked to called it XenApp 4.5, but I confirmed with her that she was referring to what everyone outside of Ft. Lauderdale calls "Presentation Server 4.5."

If you're using CATIA and willing to spend the time working with Citrix engineers on the beta (and willing to particpate in post-launch "endorsement" programs), then you might be interested in signing up for the Virtual Design Studio beta program. (They're calling this a "White Glove" program which is something that I've never heard of outside of ads for very expensive hotels. What this means in the context of CAD applications on thin clients is beyond me.)

I haven't received an update on how Virtual Design Studio will work in about six months. Last we heard, and what Citrix has confirmed publicly, is that Virtual Design Studio will leverage very special (read: expensive) dedicated graphics processing hardware that hooks into the Presentation Server to fully render the OpenGL CAD stuff on the server side. Virtual Design Studio itself will carry a price premium over the Platinum edition of Presentation Server, but all of this is chump change to CATIA customers who are paying somewhere around $15k per user already.

The big question on everyone's mind (well, on my mind at least) is how Microsoft's acquisition of Calista in January might affect Citrix's Virtual Design Studio. In a nutshell, Calista has software that lets excess CPU capacity on a host be used for a virtual GPU that's accessible to a guest VM. And while virtual GPUs don't perform nearly as well as dedicated physical GPUs, if you have 8 or 16 or 32 cores in a box, chances are you probably have a few extra cycles to spare. (And Intel doesn't mind overselling this capacity, although since AMD owns ATI, they might be able to offer some really unique hybrids in this space in a few years.)

Now it's possible that Calista will never evolve anywhere near CATIA's space that Citrix is entering with Virtual Design Studio. But Citrix has said that this is just there starting point--that they want to move into other remoting OpenGL applications and then ultimately into things like Aero Glass and DirectX. And that is hard-core Calista space.

Of course Calista wasn't actually shipping any products when Microsoft bought them (must be nice!), so really we don't have any idea when or how Microsoft will "productize" the technology they got from Calista. But it will be interesting to watch.

CVDS non-NDA.ppt

Join the conversation


Send me notifications when other members comment.

Please create a username to comment.

 I talked on Virtual Design Studio with some Citrix people at the Solution Summit. It seems that this will be an affordable solution, almost dirt cheap compared to 3D/CAD application licenses. I have heard of customers who are trying to solve the OpenGL problem with huge virtual server farms with up to 32 CPUs and HP thin clients with their propriety graphics protocol, now that will be expensive. They are using a sledgehamer to solve the problem :)
seem to be a server side graphical calculation with a rack of external NVidia video board that is mutualized and shared for graphical calculation before graphics are send to client.

FYI: A "White Glove" program means Citrix will not only supply the customer with the Beta Software , but will also supply the consulting resources to install the software in the customer's pilot environment to make sure everything is working properly. Technology like this is new, so there is no prior knowlegde within the customer itself. This way the customers gets the knowledge delivered with the code.

 The term "White Glove" is a reference to the gloves used by people handling delicate relics or equipment in their area of expertise, as to not taint or damage them. Citrix is not the only one using this term btw. 


Hi Brian:

it would be awesome if you included virtualization industry analysis on your website. We would love to hear from an expert like you on how the industry is evolving. Stuff like information on different startups, where all the hot technology is brewing.  



Yeah that's exactly what this is. Let me ask this to the community: I know that one of the problems Citrix was having in the early days of Pictor was that the Nvidia servers could only be connected to a single server (or, actually, to multiple servers). But it didn't work the other way around. You couldn't connect a bank of Nvidia servers to a single Citrix Presentation Server. Does anyone have an update on whether that's still the case? If so, does that even matter?

I have to reference my bias, I liked the product so much I went to work for the company.  I work for RES.

RES Powerfuse offers an agent within its software suite named Subscriber.  Installation of the Subscriber Agent to a Windows OS Instance within the network has the user launching a published desktop as an initial application, ICA or RDP, it is your choice. 

What is in turn subscribed into the published desktop are applications installed locally on the device.  The application icon appears in the context of the published desktop, and the user is able to launch the application within the context of this published desktop (the shell environment is uniform to the user, no surfing file trees to attach an email) local resources and all.  When I demo subscriber to customers I specifically dumb down the graphics of the published desktop, while launching the application which in turn accesses the local video card for delivery of the app.  This is a fantastic hit with the CAD crowd, and hospitals LOVE it too.

Nice item here is that the Application Services available within the published desktop now perform optimally, while the locally installed applications see a performance increase of their own.

I spent a number of years working in manufacturing, and this Subscriber Agent technology is a striking thing to behold.  It is possible to deliver a seamless User Workspace for CAD users, while also extending the SBC Application Service infrastructure out to what has been traditionally considered an impossible user community for SBC.  Interesting stuff....




So what about the files which are used in these CAD software packages? Especially if these files are stored centrally and I'm on the other side of the world over a relatively thin, high latency line... How about updates to my CAD software which is installed locally?

RES Subscriber is nice technology, but keep in mind that we are not talking about just the intergtration in the centralized desktop, the challenges are a bit more complicated in many situations...


Good point....I'm just wondering ; how big are these CAD files nowadays ? Is that in terms of megabytes, or gigabytes ? I have really no idea. The last time I used something 'CAD'dy, that was back in 1993, on an AT, and the Sydney Opera House was the biggest drawing I had at one entire 1.44 MB floppy.

There certainly is a point where centralized provisioning of the data and centralized working with that data is more important and above all cheaper than working with it over LAN or even WAN links. I can certainly imagine that even in a 10 Mbit DSL connection, loading a 200 MB drawing and constantly working with it simply does not work.

I haven't seen Callista or this technology at work yet, and I am very curious on seeing more because even 'normal' end-users without the requirement for CAD applications etc are going to ask for more soon enough. They will want the Aero interface, they will want full screen video, they will want desktoplike performance, and their desktops have GPU's more powerfull than their CPU in some cases, and that is simply still a problem on the Citrix server side....

Marcel Göertz


Hi Brian,  I really don't think it matters if you can only connect one rendering device to a server (at least at this point).  We need to understand what the bottleneck in performance is going to be.  One rendering device may be plenty for a server if the bottleneck turns out to be CPU.

The other thing that plays a role here is server hardware.  If you were going to attempt to connect multiple rendering devices to a single server you would need many PCIe (x8 or x16 presumably) slots to make this work, and there is not alot of hardware out there with this capability today.


You can connect up to 2 Quadro Plex graphics server to a single Virtual Design Studio server over their low voltage PCI-express external connect. Each Quadro Plex can have either 2 or 4 GPU's in it currently.

 The opposite is not true, because they connect over PCI-e you can't share one Quadro Plex across multiple servers. But then again you should not need to, the scalability factor at this point is 2-4 users per GPU depending on rendering needs.


My experience is that engineers make no compromise on performance.  Any lack of performance is unacceptable.  Where the data resides is a secondary consideration.  It is the unimpeded ability of the engineer to craft the CAD output which is first always as a priority.

And besides, I don't know the brand of the CAD software referenced in this post, but Autodesk rules the market.  Autodesk supports a maximum of 2 licensed users per hardware instance (it has been a while, but I believe it is hardware and not OS based).  Beyond my own bias, imho, CAD remains problematic as an application service offering on the SBC side.  Good luck though.



Sigh...luck has nothing to do with it.

Citrix is closely working with various vendors and end-users on this technology (the original software-based technology was co-developed and co-funded by Boeing). Yes, application performance is very important, but what is it worth if I have to wait for my data to be available to work with. It's like having a Ferrari in a traffic jam....

You might want to do your research about the professional CAD/CAM/CAE market. Yes, Autodesk is an important player, but to say that they rule the market...


AutoDesk likely rules the low end CAD market, but there are nowhere in the high-end market.  The high-end players are Dassault (Catia), Seimens (Unigraphics), and PTC (ProEngineer). 

 As the post above indicates is there are multiple facets to performance in CAD systems.  The rendering performance, thus user interactivity, is one component.  Another component is how long it takes to load/save the model data.  In today's world of partnering and outsourcing the end user may not be close to the data causing the load/save times to be unacceptable.  In this case a remote display technology such as Virtual Design Studio or HP Remote Graphics, may fite the bill.  It is going to be a balancing act like everything else in IT.


Everbody who is interested in this topic should watch this:

(Citrix Video - OpenGL Graphics Acceleration Demo At Citrix iForum 07)



Question one. The bigger the backend Server which is hosting the VM's, will give better graphically performance on the user's end?Also, I am going to participate with citrix's apollo program. Trying to utilize GPU performance to the client via hardware video cards for CAD.As for now, I have tested qumranet Solid Ice SPICE, which is a total software one package solution for cheap $200 a user.  The 2D CAD graphics we use them for is good. not great! but good. better than the xendesktop expereince with speed screen. Anybody else find any other packages which deliver CAD environments to the clients?


 Well after 1 week searching for possible solution, I still don't have anything that I can at least choose to perform further research. The Sun Visualisation System sounds good, but it relies on UNIX system, not suitable since I need remoting Windows Open GL apps.What I'm still curious about is the NVIDIA Quadro Plex S4 VCS, which adverts with  <strong>Advanced Remote Visualization solutions.</strong>Can you just connect it with a e.g. HP Server Platform and all apps. over accessed remotely automatically use the GPU?Or it works only if the apps are using GPU API, e.g CUDA? Anyone knows more information about this? Or even can give any URL with installation experience.thanks