The display acceleration and Metal compute hardware acceleration are two entirely separate things - using Metal for display acceleration just means it's used to present to screen (i.e. the canvas view). It should be faster than OpenGL, but between the final High Sierra beta and the public release something changed and presented some issues with the way Affinity's Metal renderer is implemented. It's hopefully something that will be addressed in the future. In the meantime, the OpenGL renderer was tweaked to compensate (it's noticeably faster in 1.6 than 1.5).   Metal compute is hardware acceleration, and is a back port from the iPad development where Metal implementation was necessary to achieve good performance. In particular, equirectangular projection absolutely flies using Metal compute, often hitting 60fps at 5K resolutions and above. Complex live filters like Twirl and other distortions should also redraw much faster. At the moment, however, it's limited to integrated graphics chips which you'll typically find either on MacBook models or the 21" iMacs.   You'll notice enabling Metal compute will use the integrated GPU, but you don't need to check "Use only integrated GPU". Photo can still use the discrete GPU for presenting to screen and the integrated GPU for Metal compute quite separately.   Hope that clears it up a bit!