-
Metal/the next FCP X re-write – and InfinityFabric
Sarah Kimberly Euschen , an editor/motion graphic pro who is very active in the higher-end internet X groups, has written a cool piece about why the new MacPro design might represent a breakthrough in desktop parallel processing. Like many, I had heard the “Infinity Fabric” buzzword in the WWDC presentation – but didn’t have much handle on what that meant specifically for video editors and motion graphics folk. Now I know a bit more.
It’s very digestible and I found it to be a really interesting look at some of what the FCP X/Metal re-write currently underway might yield.
For those interested:
https://themagicalworldofsakie.wordpress.com/2019/06/10/infinity-fabric-and-finalcut-pro/
Creator of XinTwo – https://www.xintwo.com
The shortest path to FCP X mastery. -
20 Replies
-
[Bill Davis] “CP X/Metal re-write currently underway might yield”
This has me a bit puzzled. Why should there be any rewrite? FCPX has “supposedly” been fully Metal-enabled all along. Or was that just marketing BS? Seems to me that if there is a rewrite of code going on (bear in mind its probably 10 years old at this point) then FCPX is possibly in worse shape than its competitors.
And just as an FYI – Apple is making speed comparisons based on a quad GPU set-up. That’s basically 4 of the top-of-the-line GPU cards from the current iMP line. Of course, in FCPX, that will be 4X faster. Duh! Seems like a huge expense compared with 1 or 2 Nvidia 2080 cards ☺
– Oliver
Oliver Peters – oliverpeters.com
-
[Oliver Peters] “This has me a bit puzzled. Why should there be any rewrite? FCPX has “supposedly” been fully Metal-enabled all along. “
To best leverage the new pathways that the research they did leading to the adoption of Afterburner and Infinity Fabric. Clearly those require code in the software to implement. And just as clearly, whatever used to call just the CPU or the GPU will need to now integrate calls to these new technologies for the future.
I don’t understand why this is confusing.
Adobe spent a significant time encoding CUDA processes in the quest for increased performance.
Apple is doing the EXACT same thing for Metal. I expect Adobe is still working to leverage CUDA to the max – and it appears Apple is taking time to leverage Metal to the max.
Business as usual for everyone, isn’t it?
From reading Sarah’s post, it looks like there are some fundamental processing pathway changes that these new processes enable. An FCP X code refresh seems like just common sense. How much of X’s App needs to be revised to use these is WAY beyond my pay grade – but apparently it’s quite a bit.
Creator of XinTwo – https://www.xintwo.com
The shortest path to FCP X mastery. -
[Bill Davis] “whatever used to call just the CPU or the GPU will need to now integrate calls to these new technologies for the future.
I don’t understand why this is confusing.”I’m confused, because that’s exactly what Metal is all about. Apple is already in the second iteration of Metal (current OS). A number of apps like Pixelmater Pro and Affinity take advantage of this, too. As do Premiere and Resolve. That’s what FCPX has been doing all along. Right? So what’s different now?
[Bill Davis] “Adobe spent a significant time encoding CUDA processes in the quest for increased performance.”
Not just Adobe, but everyone except Apple.
[Bill Davis] “I expect Adobe is still working to leverage CUDA to the max – and it appears Apple is taking time to leverage Metal to the max.”
All cross-platform developers are leveraging both.
[Bill Davis] “it looks like there are some fundamental processing pathway changes that these new processes enable. An FCP X code refresh seems like just common sense. “
Well, my guess is that – and my testing tends to bear this out – that FCPX has not been capable of taking advantage of the combination of multiple GPUs when installed. Apps like Resolve do. So with up to 4 GPUs available in this configuration, my guess is that FCPX is playing a bit of catch-up. Same for Motion and Compressor, I would imagine.
Note Premiere’s support for multiple GPUs in this video.
https://youtu.be/drlUWYh_U4I– Oliver
Oliver Peters – oliverpeters.com
Some contents or functionalities here are not available due to your cookie preferences!This happens because the functionality/content marked as “Google Youtube” uses cookies that you choosed to keep disabled. In order to view this content or use this functionality, please enable cookies: click here to open your cookie preferences.
-
[Oliver Peters] “So what’s different now?”
The now publicly understood architecture of the New Mac Pros, obviously.
Why shouldn’t the research and code changes Apple is building into the next X version reflect these new hardware capabilities? And as others like developer Roger Bolton noted on other boards, some of the stuff like Infinity Fabric is more tied to the AMD capabilities than specific Mac Stuff. So putting code support for that into the next X release seems like a no-brainer that will potentially benefit lots of X editors, even if they don’t spring for CheeseGrater Redoux!Creator of XinTwo – https://www.xintwo.com
The shortest path to FCP X mastery. -
[Bill Davis] “So putting code support for that into the next X release seems like a no-brainer that will potentially benefit lots of X editors, even if they don’t spring for CheeseGrater Redoux!”
I think that’s too much of an assumption, unless they are running multiple GPUs.
– Oliver
Oliver Peters – oliverpeters.com
-
Good video. Thanks for posting.
– Oliver
Oliver Peters – oliverpeters.com
-
[Oliver Peters] “Good video. Thanks for posting.
“I saw it on Alex Gollner’s Twitter, thought it might be useful.
-
BTW – I realize no one here wants to hear this, but with Apple focusing on 8K video with a 16-bit float pipeline, they are already behind the eight-ball in terms of where the industry could be headed. For example, Avid is being re-coded as 32-bit float with 16K project sizes possible. Also an uncompressed version of DNx and ACES workflow support.
– Oliver
Oliver Peters – oliverpeters.com
-
[Oliver Peters] “BTW – I realize no one here wants to hear this, but with Apple focusing on 8K video with a 16-bit float pipeline, they are already behind the eight-ball in terms of where the industry could be headed. For example, Avid is being re-coded as 32-bit float with 16K project sizes possible. Also an uncompressed version of DNx and ACES workflow support.”
I know people like the Gretzky/puck quote, but it is possible to skate too far ahead of the puck. I mean, hypothetically if 16K projects start becoming a common thing in 2023, and other NLEs have 32-bit 16K support in 2021, does it matter that Avid had it in 2019?
Reply to this Discussion! Login or Sign Up