This got me all excited, and I kinda lost my head for a bit. Looked at a couple of my systems, one w/ an RX 580 and one with 5770. All combinations of hardware + macOS 10.13/10.14 + FCPX 10.3.x/10.4.x had the "Faster" and "Better" options. Then I remembered what that selector was for, independent of QuickSync: a one-pass encode vs. a two-pass encode. So that's all you're seeing. The Faster is just making a single pass, whereas Better is first inspecting the content to know best how to allocate the bitrate when it encodes on the 2nd pass. Nothing to do w/ hardware acceleration in this case, I'm afraid...
I believe that you are right (especially the case on the cMP), thanks for all these info. I also don't think there is any hardware encoding involved. But just share what I've found (even they are not new at all). However, we still have no idea how much FCPX use OpenCL / METAL to assist the encoding. It's a known fact that OpenCL can be used to assist H264 encoding (e.g. use it to speed up the "Lookahead" operation. So, we still not sure how much (and how) Apple use the GPU during video exporting.
AFAIK, that "faster encode" is the "selector" to use QuickSync on the newer Mac. However, most likely exactly because QuickSync can only do single pass. This is why that faster encode option become the selector of using QuickSync.
So, nothing is changed, that is still single pass vs multi pass. But if the hardware can do hardware encoding, then the single pass option will become hardware encoding. I believe this is also true on the iMac Pro which has no QuickSync but use Vega for hardware encoding.
30Mbps is quite high bitrate for such simple video, as long as use the correct parameters to encode, the video should looks very similar regardless single pass or multi pass. Most likely that's why I can't see the difference.
Anyway, there is absolutely no hardware encoding yet. However, having some sort of hardware acceleration assist from the GPU is still possible.