[DocG]

For HEVC hw decoding support try a recent nightly. It was added ~ 2 weeks ago. For the future capabilities of kodi's decoder and render, use the DX11 builds, that you find in the testing thread.

Can't tell a difference between alpha 2 and the latest nightly. Unless I am missing something?

Id say it's also related to the fact that your 2x4gb/2x2gb runs in dual channel vs single channel for 1x8gb.

Most definitely.

Samsung RAM is M471B5173DB0, and if I remember correctly, these were individual sticks of the same model, i.e. not a kit. It is quite possible that it comes down to memory speed, and I believe I still have 2 x 2GB as kit of that same Crucial memory somewhere, so I'd be interested to see how it goes with only 4GB of the same Crucial. I'll do another test run when I get around to it next. I've also found two sticks of Kingston KVR16LS11/4, individual sticks again not a kit, and I may have a test run with those as well.

[Edit]
Found the 2 x 2GB Crucial kit (it's this one) and currently doing a more comprehensive test run. If you've read my previous edit, please ignore what I've said, for the time being I'll take it back. Results to follow later.

Did a quite comprehensive test run, and as suspected earlier, memory speed seems to be the deciding factor. I can tell very little difference between the 2 x 8GB Crucial kit, the 2 x 2GB Crucial kit, or even the 2 x 4GB Kingston RAM of the same model and same batch. I consistently get more skips with the 2 x 4GB Samsung, I even get skips in scences where there wasn't a single skip in any test run with any of the other modules. I rechecked the Samsung RAM, and while these are the same model, these actually do have different batch numbers. Could be due to slight differences due to being from a different batch, or could be due to the Samsung just not being as good as the others. I'd have to recommend a kit at this point, and if not a kit, at least modules from the same batch.

Nowhere during playback did any CPU core hit 100%. Even during certain scenes where there was a lot of skips, the most I ever saw was 70% on a core. I did see one or the other CPU core go above 90% in the Kazu file, but at those points there wasn't a single skip. The reported framerate jumped in a very regular fashion between two particular values, around 23.4 and 24.4, but playback was seemingly unaffected. For particular 4K HEVC encodes (at least 23.98/24 fps with reasonable bitrate) the i3 will absolutely do.