I've always made sure my imported video files were on a separate drive from the OS/Application and since I've always had the luxury of a tower with multiple drives, I also put render folders on yet another drive so disk writes of the render output don't interfere with the reads of the video data for playback or whatever.

Even with 14GB of memory, I find FCP bogs down. Anything else to be done to speed up performance other than SSD drives?

In my experience the video card has the biggest impact. I upgraded to a Radeon HD5870 and never looked back. If your pockets are deep there's Sapphire HD7950 for $500. You can also let FCP transcode your imported video to ProRes but it sure eats up hard space. My current library for a two camera shoot is over 500gb. You can also upgrade the cpu but I didn't see a huge improvement.

@Pete: Good point about the graphics card. I see that deficit a lot when I do screen switches. I currently have two NVIDIA GeForce GT 120 with 512MB cards. But the second one is just for the occasional big screen TV viewing.

Do you know, can I upgrade just one card and leave the other or must the two match?

A more useful pairing is with the hd 5770 as explained here http://www.apple.com/shop/question/answers/readonly/can-i-put-2-of-these-5870-graphic-cards-in-my-mac-pro-instead-of-two-5770-standard-graphic-cards-in-the-newest-mac-pro/Q497UAXXCYU7H7DUA

I would replace both gt120 with the 5870 which can run both monitors but you might need a mini port adapter. The 5870 go for around $170 used on ebay. Just make sure your mac is compatible, mine was a 2009 and the card must be the mac edition.

The gt120 is such a dog (I know because that's what I was using before I upgraded). Its hard to believe people still buy them for $50. The 5870 is 16x faster. You'll be kicking yourself for not upgrading sooner.

Thanks for that link. I have the early 2009 Mac Pro with two Apple HD Displays (one on display port and the other on DVI) plus the TV on the second card for a total of three. Wouldn't I get a little boost by putting each Apple display on it's own GT 120 card?

I was going to get a Saffire to run my two HD displays and leave the GT 120 for the TV. Won't that work?

There no harm in trying. Just make sure FCP is using the faster card to render and encode. I'm not sure how it picks.

The GT 120 is the cause of your problems. Since the replacements cards can drive two monitors, keeping a GT120 in the computer seems like it would be redundant. I am pretty sure that FCPX runs off the card that the program's GUI is running on. You should make sure that the program doesn't spread the computation across the graphic cards. The GT120 might cause problems if it does.

Thanks. I've done a few hours of multiclip editing using one GT 120 for each HD display and I have FCP X in two window mode with the main window spilling the inspector and effects over to the second window. It's actually much improved over having both displays on one card.

I ordered the HD 5870 which can run all three displays if I need it but will experiment with using the GT 120 for just the TV when needed. TBD

That's an excellent interim solution.

Still awaiting 16GB memory upgrade but adding a Radeon 5870 to drive the two HD Displays and leaving one GT 120 for the HDMI TV works fine. I worked with FCP a while and can feel the difference. Mission accomplished.

I believe I read that the FCP X and Compressor architecture uses GPUs for rendering. If that's true, is there any rendering performance gain in putting the other GT 120 back in with a displays? I can leave the TV unplugged if there's a gain in that too. TIA

Somebody please correct me if I am wrong, Compressor relies on the CPUs to do it's work where FCPX uses the GPUs. This is why Apple added the option of using the GPU when sending FCP projects directly to Compressor without rendering a ProRes file first. This speeds up Compressor by allowing the GPU to do what it does best and then letting Compressor render the final file in the way it does best. This is especially helpful on dual CPU MacPros.

Interesting.... I did a quick test and the difference is noticable. Here's some timings

For my completed project:

Rendering a scratch H.264 at 3K data rate took:
Send File To Compressor 9:39
Share Master File 6:46.

Using Automatic Data Rate (ends up to be 7K):
Compressor 8:39
Share Master 3:35.

Share Master File specifying the native format of the clips (XDCAM EX 35) took 2:16 plus 6:08 to manually render in Compressor to 3K H.264 for a total of 8:24 plus the time to setup the job

Share Master File specifying ProRes 422 took 1:35 plus 6:35 to manually render in Compressor to 3K H.264 for a total of 8:10 plus the time to setup the job

All this is on a 2 x 2.26 GHz Quad-Core Intel Xeon. In all cases my CPU meters look like the attached.