It's actually fully deniable. We don't know the compression methods they used for both vids, then they had to paste them both together and recompress into a single video. It is not proof of anything at all.
You need to post quality comparisons in a lossless for them to be valid, plain and simple. There is no other way to do it. Once you introduce compression, the comparison is 100% completely useless.
Now, again, it could very well still be true(and most likely is to an extent), but this video is not the proof.
As far as 10.3a boosting performance quite a bit, yeah, if you consider less than 2% in most everything quite a bit.