I have used/seen a Mac Centris 650 and a low end SGI computer, Idigo?, both
run VoxelView. With a 90 section Z series 3D reconstruction the Mac took
something closer to 60 seconds and the SGI took something closer to 3
seconds. It was impressive but cost has to be a factor, the differences are
large. I have never seen a decent PC run the above and would be interested in
the comparison. As others have mentioned, how are your Unix skills? The
learning curve verse the DOS/WINDOWS or Mac SYS/FNDR is much steeper to
really use the machine to its potential. If you expect to do a large
amount of 3D work the power of an SGI is nice.  Every filter, rotation or
color LUT change, and more, rebuilds the image you see and time adds up quick
on a slow machine. For a real time feel, something you would use to do a
presentation, power of the SGI is the only way to go. I have to think
though that a machine such as you have specified or a higher end Mac
(Quadra range) would be a good starter machine for some 3D work with a
reasonable trade-off between time and dollars.
 
 
 
On Wed, 9 Feb 1994, MARSHALL H. MONTROSE wrote:
 
>         We are submitting a shared instrumentation grant to get a zeiss
> confocal (LSM 410). As part of the equipment we are requesting
> workstations, and I need some advice on the most reasonable choices for
> both the workstation and the associated software. Our needs are 1) 3-D
> reconstruction which allows optional viewing of interior structure, and
> 2) software which allows extraction of data (pixel intensity, ROI info)
> and image calculations (ratioing, masking of ratio images based on
> intensity of original raw data fluorescence, edge detection, etc)
> conveniently on whole sets of images (eg. from time course experiments.
>         Since the SGI machines are so darn expensive, I
> am considering the following (total cost w/software <$10,000): 586/60mHz
> (pentium chip) computer with 16mB RAM, 256kB cache, 424mB hard disk,
> Windows video accelerator (Diamond Viper board). Mass data storage on
> Optical Access 1000mB read/write optical disks. The 3-D software would be
> VoxBlast (Vaytek). According to Vaytek, this setup should run about a
> million voxels/sec calculations (i.e. similar to SGI?). For data
> extraction/analysis, I was thinking of Matlab with Image Processing
> toolbox. A main worry is that this may be hard for people to use,
> although maybe someone has already built a nice GUI interface and
> packaged some nice functions together? I would appreciate any suggestions
> for better software or hardware that the NIH (and we) would find superior.
> Thanks,
>
>         Chip Montrose
>         Johns Hopkins University
>                 tel:410-955-9681
>                 FAX:410-955-9677