Issue with HP Z820 64GB RAM and dual Quadro K4000's and rendering and playback performance

Question

I'm having major issues with performance on my system (running windows 8.1. Premiere CC 7.2.2). My friend in the office here has a Mac Pro (2013) with 6 core cpu and dual D500's and I have dual Xeon 2665's and dual K4000's.
The problem is that he can export twice as fast as me on mercury software and almost 3 times faster on Mercury GPU. He can also add twice as many affects before his system drops frames (using same settings and video file). When I go to GPU-Z my gpu's are hardly working with GPU export selected... with software only my CPU's get to 60% load max whereas his CPU get's to about 84%. Am I missing something? I know the D500's are almost the same as W8000's (and he has two), but dual K4000's shouldn't be that far behind? And how does selecting cpu exporting get him an export time of 6 minutes compared to my 15 minutes when I have a 32 threads to work with??
Am I doing something wrong? Have I got a setting somewhere that is causing this that I might have missed? Are dual D500's really THAT much better than dual K4000's? AND HOW is my 7k+ machine that much worse than his £3200 mac?
Please help. I'm losing my mind here!
Thanks
Edit - Would just like to add that the video files we are using are pro res 422 and that I have the latest bios and latest nvidia performance drivers

Answer 1

That HP Z820 should kick the Mac pros butt! For example we run Autodesk Flame systems ( 150k software) on the Z820's with Quadro 6000 cards and they fly. We also had a new mac pro in for testing a few months ago w the dual fire pro cards all decked out and it was crap next the the Z820. Of cource the new MP cant run allot of highend software because of its lack of nvidia support and the firepros lack of open GL but we did run premier and smoke etc  on it for testing. ( Although i thought it looked cool the total lack of exandability and nvida support  was to much so we sent it back). Graphic workstations need PCIe expandability and a choice in GPs.
Anyhow here are a few observations:
-Software being used will dertermine the hardware purchased,
-The dual xeons at the much lower clock speed are meant to work 24/7 like a lamans endurance race, not a sprint race.  Better to get highest frequency xeons.
-Having dual CPUs with all thoughs cores will only work with a software that is properly mulithreaded, premier is definetly not one of those softwares so you not getting any advantage.
- the advantage to Quadro cards are much more noticable in 3d tha 2d. If your running Autodesk maya, 3ds max or the recently disscontinued XSI you would be flying compared to the W7000.
-dual GPU cards do not preform at twice the speed. On average the second card gives about a 10 percent boost.
-At present ( and as akways) , nividia Quadro drivers are much more uptodate and bug free compared to ATI highend cards. This will inevitably change as openCL over takes openGL, until then Quadros are the clearly superior cards for those with DEEP pockets.
-I think you would have been far better off with a single Quadro 6000 vs dual K4000s. Also any higher end ATI firepro card would have worked since Premier now uses open CL and no longer needs openGL. Main advantage to Quadro at this point (  besides 3D ) is to be compatible with the most softwares and get the best and most optimized drivers. Firepros will work properly in far fewer softwares at this point but I can see a day over the next few years this may change but not as of yet in 2014. So his W7000 wins this sprint race in this single simple example but in most any other situations your cards will actually work and win. I like saving money like the next guy but I even purchase Quadro cards for my home use and experimenting, but Im not a gamer.
-FYI: Cuda core count no longer is true measure. For example older Fermi cards worked at twice the frequency than the newer kepler architecture so in effect today you need double ( actually more) the amount of cuda cores to equal the older cards. Also the true compute power in the new Keplers is also much inferior to the older cards.  Its all creative advertiseing today more akin to creative accounting / smoke and mirrors. What im trying to explain is when people read the charts that show the cuda core count and say the higest is best is completely not true. It may have the most "cores" but it will not render correctly, double percision is missing, colors may pop, anti aliasing is inferior and inconsistant,  interaction ( most important) and shaders will be buggy or missing. There are very good reasons people spend money on highend Quadro cards but the charts and reviews generically speaking never understand or can even test for this stuff.
-Even given the above scenarios your Z820 should be killing your competition. Using a decked out Z820 is way overkill for anything CC. ( Kinda like driving an F1 to get groceries.)  get your HP rep to come over and replace / test your rig. Its free and covered in your coverage.
I am CLEARLY not a tech guy as my talent in is creation and design but I overhear the tech guys talking.