Not a dumb question at all . We offer a service called "Split Frame Rendering," wherein the farm divides your image into a grid (you specify how tight the grid should be, e.g., 5x5 or 10x10). It is supported on all of our render engines with the exception of Vue. In the case of 3dStudio, we simply use the "region render" option to mask away everything that is not in the current bucker.
There are some drawbacks that you need to consider:
1) It is *extremely* difficult to predict how long this will take or what sort of speedup you can expect. Even harder than with animations!! What we typically do is offer a free split-frame render at low resolution prior to submitting the high-resolution version. I.e., if you submit a 640x480 render with a 5x5 split, and it takes 5 GHz*Hrs to complete, then we can guess that a 6400x4800 render with a 5x5 split will take 500 GHz*Hrs, because the hi-res version will have 100 times the number of pixels. Even then, though, the cost is difficult to predict, because pixel count is not the only variable involved. See the discussion of GI below...
2) Our experience shows that it does not offer a significant speedup for most scenes that require GI. This is because with GI, the radiosity solution / photon map / etc., must be calculated regardless of which pixels are being rendered. In raytracing or scanline rendering, this isn't the case, so more benefit is derived from splitting up into smaller buckets. Steps can be taken to curtail the duplicated effort involved, although I do not yet have enough experience with Vray to know if they are viable solutions. Some examples of possible solutions are building radiosity solutions, caching shadow maps, etc., prior to rendering.
3) Our system does not yet reassemble the file. In other words, the output files we deliver is a series of images that need to be stitched together in Photoshop (or similar).
The benefit of course is:
1) If you do not use GI (E-light is great for faking GI in 3dStudio), can cache the GI solution prior to submitting on our farm, or the render engine you are using uses monte carlo path tracing to get GI instead of radiosity, you can see near-linear speedups. I.e., a 5x5 split could give you a 25x speedup, and a 10x10 split could give you a 100x speedup. Reality will probably fall a little short of that, but it will be astonishingly fast.
2) The reassembly can of course be automated. We have several customers that have written Photoshop scripts and do 5x5 splits on a regular basis since their scripts are designed to reassemble at a particular resolution with that split. They'll even submit 3 or 4 5x5 shots to run simultaneously, thereby using 100 computers in the background while they prep the next render. I've heard rumours of success using ImageMagick, but have found its scripting interface a little unwieldy for my taste.