directX option

Scott is right.
The only advantage I have to dicing content is that I can run pixel to pixel files and maximize the resolution at each projector. With the Collage generator, this is not the convention but it is something I use quite frequently on Catalysts Systems I use to run multiple projectors.
Basically the DL projectors or most standard projectors have an LCD or DLP pixel count of 1024 x 768. Therefore, the highest rez a user will ever see is 1024 x 768. If I create stock content that I need to be high rez, then I design it so that after all imaging is scaled at 1:1 that my source file is the same as my output raster which is 1024 x 768.
If for example I have a design for a show that will use 6 DL2s to create a 6x1 horizontal blend region then my screen size becomes 5376 x 768 allowing for a 128 pixel overlap between projectors. So when I'm using Maya or After Effects, etc to create my custom content then I'll create it at that size of 5376 x 768 as my source file before rendering, cutting, and compressing into the six pieces that are 1024 x 768.
This was a very complex process to establish but I've already written all of my batch scripts for rendering etc for different screen scenarios that I can pump out content pretty damned fast. (render cluster)
It actually becomes more complex than this though because I try to use standard DV codecs for my Catalyst content (720 x 480) which obviously is not to the same high projector pixel resolution. I can explain this more but I am writing a book on digital imagery content creation and I'd rather just reference it once I've released.
For now though consider this...
If I create a DL2 collage table that is 4 x 4 and I make my stock content 2048 x 1536 (3:4 ratio) for each single file then when I turn on the collage generator, each DL2 shows only a sixteenth of the file and that 1/16th is at 640 x 480 but the DL2 is still rendering and processing a file that is 2048 x 1536 just zoomed to 4 times it's size.
This becomes a processor load and totally kills performance with disk seek times making one single file completely bog down the media server if it's video playback.
In my scenario of chopping content, I could theoretically create a source content file that is 4096 x 3072 and just chop it into 16 pieces that are 1024 x 768 and the processor load is less than a quarter of the larger file but at a resolution that is twice as high.
Furthermore, even if I kept my source file at 2048 x 1536 then when I manually chop it in 16 pieces, the pieces are still 640 x 480 but at one eight the size per projector at a profoundly lower processing load.
There are two alternatives. The collage generator is fast and you put one piece of content on all machines and only suffer resolution and more processor expense. chopping content on the other hand is tedious, and you end up with tons more files to load and name as well as having to deal with keeping a naming and numbering convention that works to show savings in processing resources and an increase in resolution.

Scott is right.
The only advantage I have to dicing content is that I can run pixel to pixel files and maximize the resolution at each projector. With the Collage generator, this is not the convention but it is something I use quite frequently on Catalysts Systems I use to run multiple projectors.
Basically the DL projectors or most standard projectors have an LCD or DLP pixel count of 1024 x 768. Therefore, the highest rez a user will ever see is 1024 x 768. If I create stock content that I need to be high rez, then I design it so that after all imaging is scaled at 1:1 that my source file is the same as my output raster which is 1024 x 768.
If for example I have a design for a show that will use 6 DL2s to create a 6x1 horizontal blend region then my screen size becomes 5376 x 768 allowing for a 128 pixel overlap between projectors. So when I'm using Maya or After Effects, etc to create my custom content then I'll create it at that size of 5376 x 768 as my source file before rendering, cutting, and compressing into the six pieces that are 1024 x 768.
This was a very complex process to establish but I've already written all of my batch scripts for rendering etc for different screen scenarios that I can pump out content pretty damned fast. (render cluster)
It actually becomes more complex than this though because I try to use standard DV codecs for my Catalyst content (720 x 480) which obviously is not to the same high projector pixel resolution. I can explain this more but I am writing a book on digital imagery content creation and I'd rather just reference it once I've released.
For now though consider this...
If I create a DL2 collage table that is 4 x 4 and I make my stock content 2048 x 1536 (3:4 ratio) for each single file then when I turn on the collage generator, each DL2 shows only a sixteenth of the file and that 1/16th is at 640 x 480 but the DL2 is still rendering and processing a file that is 2048 x 1536 just zoomed to 4 times it's size.
This becomes a processor load and totally kills performance with disk seek times making one single file completely bog down the media server if it's video playback.
In my scenario of chopping content, I could theoretically create a source content file that is 4096 x 3072 and just chop it into 16 pieces that are 1024 x 768 and the processor load is less than a quarter of the larger file but at a resolution that is twice as high.
Furthermore, even if I kept my source file at 2048 x 1536 then when I manually chop it in 16 pieces, the pieces are still 640 x 480 but at one eight the size per projector at a profoundly lower processing load.
There are two alternatives. The collage generator is fast and you put one piece of content on all machines and only suffer resolution and more processor expense. chopping content on the other hand is tedious, and you end up with tons more files to load and name as well as having to deal with keeping a naming and numbering convention that works to show savings in processing resources and an increase in resolution.