Just a quick note to say I've now tested the MBP with Retina display running Resolve Lite (will install and try full later today) I had to turn off "automatic graphics switching" (as it was w/ previous gen 17") and while I still get prompted as having an error, Resolve now seems to work and I'm getting 25fps RT on a simple correct, single node. This isn't a full version but will install that and try standard candle later. BUT IT WORKS!!!!!!!! The resolution change on the retina display to 1920x1080 works really well at displaying the whole Resolve interface. Preliminary test of CUDA performance in AE of the 650M puts it way behind 480 but possibly ahead of the Q4000 (this is still running)... however this test is of just CUDA processing and may not be indicative of Resolve performance, if/when we get full support. Additionally, generally v1 of new graphic drivers from Apple tend to be very weak. So grain of salt all around.
Im really curious to hear how smoke 2013 beta runs on it, both in display and UI terms, and in performance. If it runs ok maybe ill buy one...
Benchmarked in Resolve and AE: Resolve first: Standard Candle Test result: 2 nodes = 22 4 nodes =13 6 nodes = 10 12 nodes = 5 NR 4 = 8 NR 3 = 5 NR 2 = 2 NR 1 = 1 Like a talking dog.. it's amazing it can happen at all. There could probably be a lot done to optimize performance. Looking forward to seeing if there is some mention of this model in the DaVinci config guide. As a frame of reference here is Colin Travers result from his Single GTX 580 rig: V1 2 nodes : 24 fps V2 4 nodes : 24 fps V3 6 nodes : 24 fps V4 12 nodes : 16.5 fps V5 NR 4 : 24 fps V6 NR 3 : 24 fps V7 NR 2 : 13.5 fps V8 NR 1 : 4 fps In our AE CS6 CUDA Benchmark we got a render time of 58 min 18sec... Significantly slower than Quadro 4000 which scores 42min (the Q4000 is very slow compared to GTX570/580) but still faster than the Quadro FX 4800 which scored 61min!
