Three years ago, I built a super NLE, based on Puget Systems, and now, Newegg's recommendation for "money is no object" top performing editing workstation. The result was a dual Xeon E5-2667-based system with 128GB RAM, Quadro M6000 and all SSD boot and media drives,with separate drives dedicated to projects, video and audio and temp swap areas. Fifteen thousand dollars went into that system, and I fully expected to be able to edit 4K smoothly. I was already editing 4K on a 7 year old Core2Quad system, but it would drop a few frames now and then with 4K XAVC material, and I wanted faster render times and the ability to do multicamera 4K, which the old PC wasn't up to. Fast forward to August 2015. I finally build the 'dream machine' with the fastest hardware and two hyper threading 8-core Xeons clocked at 3.2GHz. But the result? "Sticking" playback. Sometimes video would not play at all, the CTI would move and audio would play, but the video would stay frozen. Sometimes it would unstick and play a few frames then drop a bunch, then play. I found that if I stop/start/stop/start the playback, it would finally smooth out a bit. It might even play for a whole 30 seconds a 24FPS 4K timeline without a single dropped frame. But my Core2Quad from 2008 can do that too. And 60P? Even 1080 60P? Fuggetabbouttit! Over 60% of the frames drop during playback. This year, given me backup NLE/general purpose machine had crossed the ten year age boundary, I built a 'cheap' $1500 midrange system, consisting of Z390 based board, i7-9700K and only 32GB RAM. The one clearly superior aspect was the M.2 boot drive. 3X faster than SATA SSDs. And that machine cannot install Windows 7, so I was forced to install Windows 10 Pro. Well what do you know? Premiere runs like a dream on this cheap box! Not a single dropped frame. No 20-second wait for the Render dialog to open when using NVENC. Even things like GPU rendering are 2.5X faster than the dual Xeon, and I only have a GTX1060 in this new build. It will render out a minute of 4K HEVC footage in 24 seconds. Thinking that the newer GTX1060 was responsible for the faster renders, I popped for a GTX1080Ti and put it in the dual Xeon. Not much improvement over the M6000. Maybe 10-12%, but still barely cracks realtime render speeds for HEVC. The one place where the dual Xeon shines is in Maya, playing shaded viewport renders with high complexity. Oh, also DaVinci Resolve. I managed to get five 4K clips, one background, 4 PiP, rotated and all with LUTs and sharpening applied, to play smoothly at once. Only problem is DaVinci can't rendering anything useful above 1080P, so it's great for feeling the speed of playback, but useless for actual output. Funny thing is, the big boys keep touting more cores as the panacea for NLE work. My experience contradicts that.
... View more