So what's the status on hardware-accelerated rendering for video editing?

[poisondeathray]

That was a 1x1080p test , but confirmed on 2 different maxwell cards . Yes, I can confirm all those go up a few % with sAQ , at least with ffmpeg nvenc - and it makes sense that resolution would change the magnitude. There might be some variation due to OS / driver / card generation but I can definitely say it's being used more with sAQ

Yes pascal is almost 2x as fast . Crazy. Time to upgrade

I think you were bottlenecked somewhere , if you only got 3x faster than x264/medium (post#13) ? Maybe that was a typo ? Maybe I/O limited ?

[sophisticles]

There is definitely a bottleneck somewhere, I haven't had a chance to figure it out, I'm leaning towards I/O, I have a couple of 1.5tb hard drives laying around, been thinking about setting up a raid to see how that works out.

I wouldn't upgrade to a Pascal, despite my comments in this thread. I would wait until either ffmpeg or anconv support hardware vp9 encoding and then either pickup a cheap Kaby Lake laptop; alternatively I have read reports that Nvidia's next cards will support hardware VP9 encoding or maybe wait to see what AMD's upcoming Ryzen based APU's have to offer.

My next purchase will be based on if it has hardware VP9 encoding support.

[poisondeathray]

For your UHD prores4444 test, for I/O mediainfo says about 2570Mb/s which is ~320 megabytes/sec for I/O . So a 59.94 fps 1x realtime speed requires ~320 MB/s transfer at least .

In addition to I/O, there probably is a ffmpeg prores decoding bottleneck . On top of that, there is going to be colorspace conversions, bitdepth conversions if output is something other than 10bit444 YUV



Yes, I want to see how the AMD offerings pan out too . Awesome that they are back in the game