Latency-Reduction vs Performance - which RA cores are best suited to the runahead method?

[retroFuture]

I've been experimenting with latency-reduction via RetroArch's runahead method and am very pleased with the results.
Whilst a truly lag-free experience can only be achieved with original hardware and CRT displays, RetroArch produces results that are, in my opinion, good enough.
However, this technique often comes with a serious performance cost. I have been experimenting with various RA cores to see which are better suited to latency-reduction on more modest specs. Surprisingly, Genesis Plus GX has no noticeable performance loss, whereas BSNES-balanced suffers audio and framerate stutters as does Nestopia. After switching out to less accurate cores, (SNES 9X and FCEUX) I have seen a performance boost in conjunction with significant lag-reduction.

Unfortunately, I am not having any luck with Gameboy Advance, as the audio and framerate suffer from similar stuttering. My default emulator is MGBA and even the VBA core did not produce satisfactory results. I would like to compile a list of cores that support the runahead method (I believe the MAME cores do not) as well as which cores are good substitutes.

Does anyone have any RA core/setting suggestions that they've found useful as it relates to latency-reduction? (I am only referring to emulator settings, not TV/display settings.)