I've never benchmarked the old C video renderer vs the new C video renderer, so I don't really know. I did some really basic comparisons between the old ARM version and new C version and the latter was just a little slower, but these weren't tests with a lot of sprites so probably not a great example. The newer code is better at things, like rotated output, SGX support, width scaling, etc..
There's no ARM ASM for the new video renderer. I was pushing to do that and that's what kept me from releasing the Pandora version for so long. It ended up being a lot more of a headache than I thought it'd be and I got frustrated and didn't finish writing the first function.