Can anyone enlighten me as to why Blenders’ internal renderer stops using full processor power about 7/8ths into a frame? I’ve noticed that (Working at PAL res with a 4x4 matrix),
the renderer utilises 100% of all 4 cores up to part 14 and then spins down to about 25% of processor to complete the frame. There are instances where the last 2 tiles of the frame take almost as long to render as the previous 14. Surely it would make sense to utilise 100% processor power for 100% of the render. I love the internal render as the results are WYSIWYG, but it can be frustratingly slow. Not a critisism - merely curiosity 
.
I think the following happens: Blender assigns cores to X/Y Parts of the render (you can observe this by setting X and Y Parts to one - only 25% of your 4 core machine will be used). So when a part is finished, the core is then assigned to render the next one. Therefore, at some point all 16 parts of the picture are either rendered or are already assigned to a core. Now, let’s say one core finishes his work - no new renderpart can be assigned and thus it just isn’t used. The remaining 3 cores work at 75% of your CPU capacity on the rest.
Not sure if this is really what happens, but observation clearly suggests it. Maybe some expert can help to clarify.
but if so, why doesnt’ blender assign a part of the next frame to the remaining cores?
During render of complex scenes with ultra high vert counts, etc., this would surely cause Blender to crash.
Why ask why? Try Kind-Bud, dry!