I'm not sure if the Mac Pro section is really the right one or if I should punt this to the Video Editing section. Bear with me while I text-vomit out various thoughts regarding some of the latest announcements from Apple and Adobe. This'll get a little long-winded, so have a cup of joe handy while you read it. Tutor and I have been chatting about this via PM for the last week or so. Hopefully he doesn't mind me making it public so that I can get some other eyeballs and thoughts on it. Knowing him, I doubt he'll mind whatsoever; bear in mind I'm not belittling his suggestions in the least regarding video cards (summary: he said stick w/nVidia. That'll make more sense in a moment). First and foremost: I think Apple's new Mac Pro looks like a great piece of machinery. But I don't want it and am not going to buy it. Instead, as I've mentioned in Tutor's "CPU Performance" thread, I'm going to wait for the Ivy Bridge Xeons to be released and build a dual-proc Hack. It's likely that'll require waiting for 10.9 Mavericks to also be released as the present OS X probably doesn't have the necessary unlocks for those chips. Tutor has already pointed me at a suggested motherboard (when it becomes available...). It'll require a specific type of case as it's not an ATX-sized board. The challenge is: Which video card or cards? I've always been an nVidia fan. My current modified Mac Pro 5,1 has one of Dave's flashed GTX570s in it, and I use it quite a bit with Adobe's Premiere Pro (CS5.5 and CS6). Apple is making a clear statement that OpenCL, not CUDA, is the wave of their future. Adobe acknowledged that last summer when they released CS6 on the Mac with both CUDA and OpenCL support. We know nVidia cards can do both, but so far they've been very weak with OpenCL. Is that a driver thing specific to the nVidia cards? Is it a more generic driver within the OS? Will we see an improvement in OpenCL performance with nVidia cards and drivers? Or, perhaps, is it the actual hardware on the cards? Are AMD's cards better suited for OpenCL than nVidia's? The other interesting variable to bear in mind: the NLE. Premiere CS5.5 only supports CUDA and will never be updated again. Premiere CS6 does support both, but from now on will only be patched to fix bugs (for however long Adobe deems). Neither version can address multiple GPUs and never will. Worse, CS6 has a nasty AVCHD bug (I shoot in all AVCHD) making it utterly unusable to me until Adobe patches it... if they ever patch it. Their new CC version of Premiere has fixed the AVCHD issue, and has better OpenCL support, and supports multiple GPUs. But, it's rental-only software, and I'm not interested in signing for that. It was almost perfect, too... Then there's Apple's Final Cut Pro X. Personally, I don't have a problem with their big UI change and have been able to use it quite easily. It only supports OpenCL-enabled cards, but it can also make use of Intel's AVX extensions. AVX is something that was introduced with Sandy Bridge; my "old" Mac Pro's Westmere's don't have it. My new Macbook Pro Retina does. What I can't find from Apple or any of the user forums is: what specifically does FCPX throw at the GPUs for acceleration? Adobe makes it very clear what they can and can't do with GPUs; the lists are out there. But the same lists for FCPX don't appear to exist. How much faster is FCPX with a good card? What can it do with a good card? What IS a good card for it? Here are some interesting homemade benchmarks I did with a couple of my machines and the various NLEs: I have a 20 minute AVCHD clip that I import into the NLE, edit it, then export it to 720p h.264 MP4. Assume time t is the length of the footage. With Premiere Pro CS5.5 and Dave's GTX570, the export is 1/2t. CS6 is a non-starter since I can't edit AVCHD footage with it. FCPX on the same machine: 2t. And with FCPX specifically: I did not transcode to ProRes upon import; I wanted to speed things up. Just to make things a little interesting, I installed FCPX on my Retina MBP and cut the same chunk of AVCHD footage. It performed the export in time t. Basically: twice as fast as my Mac Pro. What's interesting is I had the GT 650M in the laptop disabled (accidentally) when I did the test. I think the speed-up may have been due to the AVX extensions, but I'm not 100% positive of that. Lots to digest, and I appreciate that you're still with me here. Some would say, "Apparently CS5.5 works for you and is fast. Stick with it!" I would be glad to, but while playing in the present, I'm watching the future. Adobe has already promised they'll make sure CS6 works in one more future version of Apple's OS (ie: Mavericks). After that, they're done patching it. CS5.5 is basically dead to them. And of note: I already know that After Effects CS6 won't work in Mavericks until Adobe patches it (which they're working on). The same may be true for Premiere, but I don't know yet. Basically, I think staying with the CS* train of Adobe's software is a dead end if I want to use newer hardware and operating systems. That leaves me with Adobe's software rental or the uncertainty of FCPX's performance with various pieces of hardware. Of importance: I do not and never will game with my Mac Pro. I have a Windows gaming rig for that (which will always have an nVidia card or two in it). I'm less set on nVidia for my Mac Pro and will move to AMD's boards if it makes performance sense to do so with my NLE.