I'm more familiar with nVidia cards than AMD, so this may sound like a basic question, but why do you need both cards? Is hardware encoding only available on the RX 570?
At any rate, unless Twitch is giving you transcoding, I'd stick to software encoding for now. The main benefit to software encoding is that you can easily tweak it to be more bitrate efficient, while GPU encoding tends to offer less control over encoding speed (usually working at an equivalent to software encoding at ultrafast or veryfast), so it tends to require far more bandwidth for the same quality. The main benefit to GPU encoding is that it tends to be a lot lighter on system resources than software encoding for its quality, but again, that's only useful if you are only putting it on YouTube and don't mind every hour of video taking 2-4GB (as opposed to ~1GB) in order to have decent 1080p video quality.
As for Crossfire, I'm not super familiar with it, but I would be surprised (and impressed) if you could use it with two different cards without issues. I would expect, however, that it would either pick one of the cards to make use of, leaving the other to idle; or use both cards, but throttle down the clock and memory access speeds to match the slower card.