Suggestions on FCPX performance

[Joe Marler]

[Shawn Malone] “The reason I went with this MacPro is that the PCI slots were appealing, and my peers all said it will perform great compared to more recent Macs, with the added expandability being a plus…but I’d be willing to investigate more on newer systems for the future. If I go with an iMac or newer trash can MacPro, what is the GPU situation? Does the better CPU negate having to have a massive GPU? “

Unfortunately your peers were apparently unaware of the importance of Quick Sync or the cumulative difference in Instructions Per Clock of an 8-generation-old CPU vs a modern one. A 2017 i7 iMac 27 would be much faster on H264 material.

Re expandability I have had three 32-terabyte Thunderbolt RAID arrays plugged into my 2017 iMac simultaneously, and those work fine. Likewise when plugged into my 10-core iMac Pro. The Thunderbolt Macs have very good I/O expandability.

However the 2009 Mac Pro is still a capable machine but you will need to use proxies or optimized media to compensate for the CPU shortcomings.

The trash can Mac Pro isn’t vastly better — I had a 12-core D700 version and tested it extensively vs a 2017 iMac 27, and it was generally slower on H264 material, whether transcoding or scrubbing the timeline. It was very quiet.

GPU performance isn’t a magic wand which makes everything run faster. In general it is only usable on effects, and only those which are (1) GPU enabled and (2) GPU limited. E.g, some effects are very compute-intensive, yet use relatively little GPU. Good examples are Digital Anarchy’s Flicker Free and Imagenomic Portraiture. Neat Video noise reduction can use GPU but it’s still slow. It at least has a built-in sizing tool which allows selecting a mix of CPU cores and GPU or both, but in general using 100% GPU on Neat Video is not hugely faster than a modern 10 or 12-core CPU.

If you got a pre-owned 2017 i7 iMac 27 with Radeon Pro 580 GPU it would probably be much faster on your H264 material. Just don’t get one with a 1TB Fusion Drive. It can be a bit noisy when transcoding, but if you have stacks of RAID arrays you’ll never have a totally silent workplace.

Another option and fairly inexpensive if on sale is the base model 8-core iMac Pro. It doesn’t have Quick Sync but uses AMD’s similar UVD/VCE hardware. It is much quieter than the i7 iMac.

However it seems likely that Apple will upgrade the iMac fairly soon so it might be best to wait for that.

[Bret Williams]

It’s a mess. And of cuourse now that your project links to copy 2,3 or 20, you have to keep that in your browser or back it up with the project, and so on.

_______________________________________________________________________
https://BretFX.com FCPX Plugins & Templates for Editors & Motion Graphics Artists
Hang Tag https://bretfx.com/product/hang-tag
Overshoot Text https://bretfx.com/product/overshoot-text/
Outliner https://bretfx.com/product/outliner/
Clock Maker https://bretfx.com/product/bretfx-clock-maker/

[Shawn Malone]

Holy macaroni…switching back to the motion templates folder!

Again, thanks for the help!

Shawn

[Shawn Malone]

Joe, thanks for all this great info and comparisons. All very useful and important stuff that I can use when I’m in the market for a new setup.

I think what my friends were saying was that the 2009 MacPro I was looking at could generally perform close to the newer Macs and wouldn’t be a slug…also none of them are video editors, but they are Mac enthusiasts, so I don’t think they have any of that detailed knowledge related directly to handling video. And I am very happy with the Pro, but I didn’t have the detailed CPU/GPU info you’ve given me. Very cool to know! I did know about specific GPU enabled FX and that it sometimes it is not doing much. I don’t really use/own any GPU intensive FX packages. I just don’t need them yet. I have some extra transition and title plugins, that’s about it. All of my work as of now (and for the last year since I got the Pro) is simple stuff that is all relatively the same episodic style/content that the client likes to keep unrefined and fairly basic. It only requires a lot of different transitions and titles to spice it up…so no looks, styles, or compositing and I never have to do any noise or stabilization stuff. Not to say that that will never be the case…just not required right now.

So it appears that I have to get the iMac factory configured with certain GPUs. Correct? Do you ever feel restricted by GPUs that are available with the systems? Or does the CPU efficiency just make up for any shortcomings?

For now, I’m going to roll with the MacPro and optimized media. That seemed to work like a charm…and now that you’ve planted that seed in my head about a newer setup, I’ll be watching for some deals! But I feel much clearer about what’s going on with the system side of things…very cool!

Thanks again!
Shawn

[Doug Metz]

One other thing that may help you eek out a little more performance – the drive bays in that MacPro are only SATA II, which will limit the speed of your SSD boot drive (it sounded like that’s where you had it). OWC sells an inexpensive PCI carrier for the 2.5″ drives, and the PCI slots are faster.

Doug Metz

Anode

[Shawn Malone]

Thanks for the suggestion Doug. You are correct, my OS drive is in the motherboard SATA port. My other SSD is in an Apricorn Velocity PCI card. I planned on getting larger SSD drive (1TB) swap it for my PCI 500GB and make that my OS drive which is only a 160GB. I had also kicked around the idea of getting another PCI card to do just what you said. The Velocity does have another SATA port also for another drive. I thought about trying that out first before I got another PCI card. Thanks for the reply!

Shawn

[Pat Sitton]

I realize that this is an old thread but, to my surprise, the previous posters left out a vital factor.
FCPX is written to use OpenCL-based GPUs (such as AMD/Radeon) rather than CUDA-based GPUs (Nvidia).
As a result, there are many tests that show that cards like your GTX 980ti — while blazing fast for gaming and with Premiere Pro (which is optimized for CUDA GPUs) — are pokier than far cheaper OpenCL cards when used with FCPX. This is most likely why you found FCPX to be less responsive than PP. I don’t have URLs readily at hand for such tests but this is general knowledge and you should be able to find such discussions easily.

all the best,
Pat