Long GOP formats, or really any interframe codec, can be immensely taxing. In order to show frame 14 (lets say) the computer has to process out frames 1-13. That mean 14x the processing is required vs an intraframe codec (like prores or dnxhr) which only has to draw the required frame. In the case of say HD image with reasonably powerful gear, this can happen in a fairly effortless way. But for UHD images, that’s 4*14 the processing required for the equivalent HD frame in a intraframe codec. Or up to 56X the processing required. Additionally, AVC based codecs even on I frames (intact frames where the frame doesn’t depend on previous frames) are still more computationally taxing than the same raster in ProRes or DNxHR.
All of the decoding for those codecs is done on CPU inside of Resolve. Other possible bottlenecks; RAM, speed of ram, bus speed, bus speed to GPU, bus speed to drive, etc., etc. Dual processor systems have more total bus lanes for everything than single cpu systems. But they also generally have more raw processor power, which I suspect is your current bottleneck Dave. Of course, once you go dual xeon cpu, you’ll need more multichannel ram, and then you’ll start to hit the wall of that single Ti, so you’ll need two of those, and then you’ll start to saturate that SSD, so you’ll need to switch to 5-10 SSD raid, and then….
Gotta keep chasing that dragon.
I sent this video to BM support, here is what they said (encouraging):
“Thank you for sharing this with us. Although Resolve plays much more nicely with ProRes, DNxHD and uncompressed formats, we continue to expand its support for more compressed codecs. I see this continuing with each update to Resolve, so I’ve forwarded this to our engineers for review.”