Two points:
(1) In the very old days, .com files were supposed to start up a little quicker than .exes, but as Salem points out, this hasn't been true now for many years; in any case, start-up speed should be quick enough to be irrelevant (any comments on .net, anyone). If start-up speed is genuinely causing you problems, you need to think about what's going on. For instance, it's not a good idea to call an exe 47 million times in a loop from another application.
(2) Running speed depends on your code. Don't expect massive speed-ups by moving to nasm. If there's an underlying problem with choice of algorithm, any amount of careful optimization in assembly won't make things much faster. It's incredibly easy to write assembler that runs within 5% of the speed of compiled code, and not at all difficult to write assembler that's somewhat slower than compiled code.