[vtkusers] VTK 5.6.0 vtkGPUVolumeRayCastMapper very slow?

[leandro]

I bring this topic back to life because i had similar problem wich didnt get
100% solved with this suggestion.
The good thing is that I got my 512 x 512 x 194 (short) volume rendered in a
good quality and performance.

But I still have problems with other, larger datasets:
the first: 512 x 512 x 1359 (short) volume. It is so slow that you cannot
even call it "interaction". 
the second: 512 x 512 x 1775 (short) doesn't even render (crashes).

My system is similar to Elvis Chen's:
VTK, rebuilt usign DirectX SDK from June 2010 as proposed here.
Win 7 64-Bit
Xeon 2,7GHz 4 core 
4GB RAM
2 x NVIDIA Quadro FX 3700 with 512MB video RAM

I've been wondering, if VTK doesn't take use of both GPUs together and the
512MB video RAM of one is the bottleneck.
I wrote a pure-VTK small test-program to render a volume. It means
dicomReader -> mapper -> volume -> renderWindow.

I observed the behavior of VolView 3.4 while rendering and interacting. Even
with option "Use GPU rendering" unchecked I see alot of GPU activity (>90%)
while interacting (watched using "Process Explorer"). This doens't happen
when I load the same volume with my application (2-3%). What does VolView do
different?

I would appreciate any idea on how to explain this behavior. Thank you =)

--
View this message in context: http://vtk.1045678.n5.nabble.com/VTK-5-6-0-vtkGPUVolumeRayCastMapper-very-slow-tp1249233p5056223.html
Sent from the VTK - Users mailing list archive at Nabble.com.