New Mac Pro or new Video Card?

[surfmadison]

Previously I had posted about long renders and share times in FCPX. With a few adjustments I am saving a bit of time, but it is still no where as fast as FCP7 due to the ability to export to a QT Reference File. Plus my share files before I encode in AME are huge (Gigs instead of Mbs).

With 10.1 out and the new Mac Pros out I was thinking that a new rig would speed things up. However, I wonder if just a new video card would help.

I currently have a Mac Pro 3.33 Ghz 6 Core with the original video card (whatever came with them 3-4 years ago). With the move to OpenCL (which I don't really understand how it works), would getting some new and faster video card help with renders, shares, and encodes?

Please let me know your opinion. I shoot in HD, minimum 1 hour long videos, which get encoded down to 100 MB files for Internet distribution. We shoot a lot of video, so saving time would be of great help, short of moving back to FCP7.

 

[ChrisA]

...
Please let me know your opinion. I shoot in HD, minimum 1 hour long videos, which get encoded down to 100 MB files for Internet distribution. We shoot a lot of video, so saving time would be of great help, short of moving back to FCP7.

I'd seriously re-think your work-flow. If you are shooting in 1080p (HD) FCPX will edit in your native format with no conversion to huge file sizes. You can stay in h.264 full time. You can edit in proxy format too with is a small down sampled file format that is faster. Then last FCP will render the web-sized files directly from your camera originals.

Changing out hardware will make you your current workflow go faster by only maybe at most 1/2 the time. but changing the workflow might give you a much greater speed up.

As an example I can on my quad core iMac import 1080p foot age and begin editing right away then when I'm done click "share" and yes then it could take some time. But that is not time I have to wait for, I'm done and working on something else.

 

[nateo200]

Great thing with FCP X when you have a decent processor (your system should be able to handle it) is you can edit H264 without rendering...its not as smooth as ProRes but for simple cuts you can disable background rendering and on export if you added any effects it will render as it exports. A new GPU will make rendering super super fast, I feel like I keep repeating the same thing over but look at Nvidia graphics cards like the GTX 680...there is a guy who sells them on here ready to go for Mac Pro's and the 680 is such a great card for the price and whether you working in FCP X or rendering out a complex 3D design from Maya its going to seriously help. A new GPU will be cheaper and depending on your needs last you a long time.

One thing that scares me is the thing you said about 1 hour of video going to 100megabytes....thats wayyy to small...minimum bit rate I would upload to YouTube is 3mbps and I usually upload 6-15mbps, google bit rate calculator for more details. Also when I render out of FCP X I render out ProRes first then I use handbrake to encode H264 since handbrake will use every ounce of CPU and RAM available and run circles around compressor/FCP and rendering out ProRes is much much faster and if needed gives you a high quality master copy to create other copy's or whatever.

 

[jasonvp]

I currently have a Mac Pro 3.33 Ghz 6 Core with the original video card (whatever came with them 3-4 years ago). With the move to OpenCL (which I don't really understand how it works), would getting some new and faster video card help with renders, shares, and encodes?

I think what others have mentioned about changing your workflow are all good suggestions, and ones you should pay attention to. However, while focusing purely on hardware, you're facing a couple of challenges here with FCPX and your "old" Mac Pro:

• No AVX support in the CPU
• Poor official AMD card choice for OpenCL (7950)

FCPX takes advantage of the Intel AVX extensions in Sandy Bridge and newer CPUs. AVX extensions help speed up (sometimes significantly) floating point calculations. The chip you have in your Mac Pro doesn't have AVX available to it; there are no Westmere and prior chips that do or ever will. So, there's some performance benefits you're missing out on by staying with the "old" Pro when it comes to pure CPU power.

On top of that, FCPX can parallelize some of its calculations and feed them to the GPU or GPUs, if more than one are available. It does this via the OpenCL APIs that are available in OS X. As of now, AMD GPUs just perform OpenCL work better than nVidia ones do. I suspect that'll be the case for the foreseeable future, because nVidia would rather people code to CUDA (a competing API). Given Apple's stated move to OpenCL as of FCPX, if you're going to edit with it, it makes sense to have an AMD GPU.

If you're editing with FCPX 10.1, it makes even MORE sense to have multiple AMD GPUs. And this is where the old Pro falls short. Without adding external or drive-bay power supplies, you're not going to be able to power more than 1 modern-day AMD GPU in an old Pro. Further, the only modern AMD GPU I'm aware of that is officially supported on the Pro is the 7950. It's not a bad card, all things considered, but it's a generation or two (or more) old.

So, given that: if you're going to edit heavily with 10.1, and can afford the scratch, I'd say move to a new Mac Pro. The old ones are fantastic work horses, but they're: OLD.

 

[ChrisA]

About hardware,.... I'm seeing much better performance than you describe on a 13" Macbook Pro. Likely because of the added CPU instructions and the newer graphic card. I edit the MP4 files that come off the Canon camcorder.

 

[Donar]

...
Further, the only modern AMD GPU I'm aware of that is officially supported on the Pro is the 7950. It's not a bad card, all things considered, but it's a generation or two (or more) old.

The 7950 is based on the "Tahiti" architecture, the D500 & D700 are based on Tahiti too (D300 - Pitcairn). If i understand the thread(s) in the Hardware/MacPro subcategory right. So i wouldn't worry about that.

But as you stated it is necessary to differentiate.
If someone needs the power of two GPU's without any hassle and money is not a problem then the nMP is maybe the best choice.

If it would be just "nice to have" faster render times, then upgrading to just one 7950 seems reasonable. And if you don't mind tinkering, put two 7970 in the original MacPro with additional external power...

 

[dollystereo]

Put 2 AMD 5770 inside your machine. your performance will increase considerably for playback and rendering.
From barefeats.com:
" The 5770s require only one power feed each so we used the two factory standard PCIe power feeds. A pair of them is definitely faster rendering in FCPX 10.1 than any single AMD GPU including the Radeon HD 7970."
and is cheaper, they are sold on ebay for nothing.

 

[mBox]

Not sure if Id ever pit QT Ref with an actual codec crated file.

Thank god QT Ref sucks on AVid cause Id never trust that POS app to export anything important.

Has there been any progress with 10.1 as far as speeds go for you?

 

[rx7dude]

Great thing with FCP X ...
Also when I render out of FCP X I render out ProRes first then I use handbrake to encode H264 since handbrake will use every ounce of CPU and RAM available and run circles around compressor/FCP and rendering out ProRes is much much faster and if needed gives you a high quality master copy to create other copy's or whatever.

Thanks for the tip. I never considered using Handbrake. This is way faster.