VMD-L Mailing List

From: John Stone (johns_at_ks.uiuc.edu)
Date: Fri Jan 14 2005 - 11:29:36 CST

Thomas,
  The display configuration you're asking about is precisely what
we have setup on all of the high-end stereoscopic workstations in
our research group. At present we're using Sun workstations with
dual framelocked XVR-1000 video boards and the Stereographics
CrystalEyes glasses. I've looked into how one might do this on Linux/PC
hardware, but thus far I haven't found anything that's truly functionally
equivalent to what we can do on the Sun sparc machines, or like we used
to do with SGIs. The closest solution I can think of off the top of my
head would be for you to try one of these ideas, none of which are quite
as good as the Sun/SGI way of doing things, but they might be ok for your
purpose:
  0) Use a Linux PC with dual-PCI-express /w 2 Quadro FX cards, or something
     similar. I don't think anyone has done this yet, the hardware is just
     being announced lately, and I have yet to see these dual-video
     motherboards personally, but when they become available, this is the
     closest thing you're going to find to a real Sun/SGI workstation, in
     terms of multi-display stereo work. If anyone's got such a machine,
     I'd love to hear about it.

  1) Use a Linux PC with a stereo-capable card such as an
     NVidia Quadro FX 4000. If you limit the display resolution, that board
     _might_ be able to drive two stereoscopic displays at once. You'd be
     sacrificing things like antialiasing and such due to the massive video
     memory requirement, but if you found a video mode combination that would
     allow this, you'd have two framelocked screens (by virtue of being driven
     by a single board with a single main clock) and thus no annoying flicker
     on the "GUI" screen. If you get it to work, then you would probably also
     be able to setup X to treat both displays as a single huge desktop,
     making it very easy to work with, moving GUIs around, etc.
     
  2) Use a Linux PC with a stereo-capable card such as a Quadro FX 4000 for
     the stereo display, then use an inexpensive PCI-based 2-D video board
     for the GUI display. This can be really annoying to get setup correctly
     with XFree86, but if you manage it, you'll end up with two X11 displays,
     one at :0.0 and one at :0.1. I coded VMD to allow you to direct the
     GUI windows to a separate X display from the OpenGL window. You can
     set the DISPLAY for each like this:
       setenv DISPLAY :0.1 (GUI windows go to 2-D display)
       setenv VMDGDISPLAY :0.0 (OpenGL window goes to NVidia Quadro)
     The main disadvantage you'll have with this approach is that you won't
     have your 2-D GUI display framelocked with the main stereo display,
     and so you'll end up seeing a lot of annoying flicker on the GUI display.
     
  3) Use a Linux PC with any dual-display video card you want, so long as
     it provides VMD with a stencil buffer ("STENCIL" listed in teh VMD
     OpenGL startup messages). So long as you've got a stencil buffer, you
     could use a line-blanked stereo mode, such as the one provided by the
     old Eye3D 4-in-1 stereo glasses. This doesn't require any support from
     the host operating system or opengl drivers at all, the code is
     built-into VMD. You just need a stereo emitter/convert that can do
     line blanking and you're in business. You could wire up the line-blanked
     emitter for your stereo display, and leave the other display normal.
     The main disadvantage of line-blanked stereo is that it will drop
     the vertical resolution to half what it normally is, so you'd want to
     use a video mode with a very high vertical resolution, something like
     1920x1080 or higher. This is the cheapest of the options, and it is
     easiest to make work with almost any video card, the main issue these
     days is getting your hands on one of the Eye3D 4-in-1 emitters that
     supports this line-blanked stereo mode.
   
  4) Consider using different hardware/software instead of a Linux PC.
     There are still many good reasons to buy SGI and Sun workstations,
     stereo is one area where they are still MUCH more mature than any of
     the competition. There's a lot to be said for using the right tool
     for the right job. In our laboratory, the stereoscopic workstations
     are the last bastion of traditional high-end commercial Unix machines.
     We're now using various Linux PCs on people's desks, but for our
     stereo needs we still use Sun machines because they provide an ideal
     solution to the problems I've described above. Sun and SGI machines
     will allow you to run several video boards, often with multiple output
     channels per board, in stereo, with framelock cables that sync the
     video boards so you don't get any flicker in the "GUI" screens etc.
     The SGI and Sun machines typically have a lot more framebuffer memory
     (we have 3 year old machines with 384MB of video memory per-board,
      and a new one with 1GB per-board) which makes them VERY capable
     for viewing volumetric data, performing multisample antialiasing,
     display list caching, and other things, all without giving up stereo.
     If you "want it all" these are probably still the best solution for
     stereoscopic needs. I continue to lobby the PC graphics board vendors
     for more powerful cards, adding the key features that they lack versus
     their Unix workstation counterparts, and they continue to make progress.
     A lot of the limitations of PCs still stems from the system bus, and
     the fact that they only have a single AGP/PCIX slot, when/if they
     solve this on common motherboards, PCs will be much more Sun/SGI-like,
     and it'll be easier to do this stuff. Who knows, maybe Sun and SGI
     will have Opteron/Itanium-based machines in the middle-range price
     zone that will run Linux with multiple video boards much like their
     classic SPARC/MIPS machines. I've definitely told the Sun people I know
     that I'd love to see a new dual-video Opteron box at some point, we'll
     see what they come up with. That'd give you the option of running
     Linux or Solaris, the speed of the Opteron, with the graphics capabilities
     more similar to their classic SPARC machines. We'll see how this
     progresses.

I hope this helps, let us know if you have more questions.

  John Stone
  vmd_at_ks.uiuc.edu

On Fri, Jan 14, 2005 at 09:41:13AM +0100, Thomas Hedegaard Pedersen wrote:
> Hi,
>
> We would like to take advantage of the 3D capabilities of VMD...for that
> purpose we intend to buy a new workstation (PC with linux) with two
> monitors - one for the 3D (stereo) viewing and another one for the
> control windows.
>
> I have had a look at previous postings to the mailing list and done some
> searching on the internet, however, I'm still a bit confused about what
> equipment to buy. As far as I can see there are four reasonable
> solutions (when not considering 3D monitors!):
>
> 1) "CrystalEyes Workstation" from StereoGraphics (shutter technique)
> 2) "Monitor ZScreen" also from StereoGraphics (polarization
> technique)
> 3) "E-D 3D" from eDimensional (shutter technique...seems only to
> support page-flipping display mode...I guess that's not supported in
> VMD!!!)
> 4) "Eye3D" by i-Art (shutter technique)
>
> The latter two options seeming to be the least expensive ones by far!
> Can anyone recommend one over the other of these solutions (or other
> solutions if they are better) for the dual monitor setup????????
>
>
> Concerning graphics cards, my impression of the postings is that one
> should go for an NVIDIA card...especially a Geforce 6 series card e.g.
> 6800GT (PCI-Express or AGP???)...is that still the recommended one or
> should one go for another card e.g. an NVIDIA quadro series card (i.e.
> does VMD utilize any of the quadro features; e.g. antialiased points and
> lines, logic operations, clip regions, hardware-accelerated clip planes,
> two-sided lighting etc.)????
>
> I would very much appreciate any help on these issues, since I'm
> completely inexperienced regarding all this 3D equipment!
>
> Thanks a lot :-)
>
> --
> Thomas Hedegaard Pedersen
>
> ----------------------------------------
> MEMPHYS - Center for Biomembrane Physics
> Physical and Biophysical Chemistry Group
> Department of Chemistry
> Kemitorvet
> Building 206, room 243
> Technical University of Denmark
> DK-2800 Kgs. Lyngby
> DENMARK
>
> Phone number: +45 4525 2018
> Email : thp_at_kemi.dtu.dk 

-- 
NIH Resource for Macromolecular Modeling and Bioinformatics
Beckman Institute for Advanced Science and Technology
University of Illinois, 405 N. Mathews Ave, Urbana, IL 61801
Email: johns_at_ks.uiuc.edu                 Phone: 217-244-3349              
  WWW: http://www.ks.uiuc.edu/~johns/      Fax: 217-244-6078