i ran across what has to be one of the most thorough tests i have ever seen with regards to encoding quality between x264 and various gpu based encoders:

http://www.behardware.com/articles/828-1/h-264-encoding-cpu-vs-gpu-nvidia-cuda-amd-str...-and-x264.html

i do have a problem with some aspects of the testing, for instance they only used apps that use the reference nvcuvenc h264 encoder, main concept's cuda encoder is significantly better. also, they make a number of factual errors as well as some of their choices with regards to source material left me scratching my head.

while they conclude that software based x264 encoding is superior to gpu based encoding, they do comment that x264 on ultra fast sucks:

While ‘ultrafast’ mode is truly bad for your eyes, quality rises quickly. In ‘faster’ mode, the quality is already higher than the best of what we’ve seen up to now.
it seems that they conclude that x264 doesn't beat the quality of the gpu powered encoders they tested until the "faster" preset is used, but as is expected it's also slower than gpu based encoders.

it's also interesting to note that they also say:

What is much more interesting however is that with two passes there’s a significant improvement in quality, with veryfast 2p already significantly better.
this leads to the question if gpu powered encoders supporting 2 pass encoding (to the best of my knowledge main concept's cuda encoder does support 2 pass vbr encoding) were thrown into the mix would the quality results have likewise shifted to significantly better image quality for cuda?

all in all, an interesting, if flawed, test.