Andrew Richards
Forum Replies Created
-
I was chatting with my editor friends at a New Year’s party and FCP7 is very much alive an kicking all around DC. It sure looks like back in the day when there when all those old Power Mac 9600 Avids plugging along as FCP3/4/5 was coming into its own.
I thought by now we’d at least know which way the wind was blowing.
Best,
Andy -
The 8-core seems to be the best option for balancing highly-optimized multithreaded tasks with the still all-to-common single-threaded ones. I still think given the price of the 8-core upgrade ($1,500 over the 6-core) that the best bang for your buck is going to be the 6-core with the D700 upgrade. That said, if it is going to be used all day every day on paying gigs, that $1,500 upgrade to 8 cores for the best all-around performance should be much easier to swallow. That’s only $125/month in the first year to have a great many (most?) tasks go 25% faster, so it kind of pays for itself under the right circumstances.
The more obvious takeaway is that there is no reason to skip the top GPU option, especially as we can anticipate that to be where Apple is pushing the performance envelope with future software. I’d be surprised if they don’t announce new framework improvements in OS X at WWDC14 that make it easier for developers to build their code to support the multiple GPUs in the Mac Pro for stuff like Core Image, Core Animation, and so on. As noted in the AnandTech review, today’s apps have to specifically call the dedicated compute GPU on their own as there is no OS-level GPU balancing happening in Mavericks.
Best,
Andy -
Anand went into a fair amount of detail on how FCPX utilizes the CPU vs the GPUs and included a handy map of the distribution of busses across the TB2 ports to get the most out of the available bandwidth. His review also indirectly answered a big disappointment of mine about the new Mac Pro; no built in 10GbE NIC. There just weren’t enough PCIe lanes left in the system to support one without sacrificing a TB2 bus.
Best,
Andy -
[Jeremy Garchow] “It hasn’t been updated in a long time. It’s my connection to legacy apps. ;)”
The irony! It’s killing me!
Best,
Andy -
[Bob Zelin] “While most modern shared storage systems are now supporting NFS, shouldn’t Apple have an “official” way to do this, or simply support their own SMB2 protocol?”
They are filtering based on filesystem type, and I don’t know if they just forgot to block NFS since it is obscured on OS X or if there is some structural difference between NFS and SMB (or AFP for that matter) that would explain why NFS is allowed but the others are not. I can’t imagine why they would block the use of AFP or SMB if it weren’t for some legitimate technical reason. The best I can come up with is that the Core Data SQLite data stores that FCPX uses have very strict byte-range locking requirements, like those described in Apple’s developer docs. The puzzling thing is they mention AFP alongside NFS and HFS+ as being able to support Core Data SQLite stores. Perhaps they assume (correctly) that the typical AFP share on the typical network is not going to have the performance FCPX wants in order to work satisfactorily. Saving the users from themselves, so to speak.
There may be some key structural difference in how byte-locking works with NFS compared to AFP which explains the discrepancy. Perhaps Mr. Modica can shed some light on that; I don’t know the nuts and bolts of these protocols like he does.
Best,
Andy -
[Bret Williams] “But what about the 3 years of OS and App updates? That clone would be completely irrelevant after 18 months. After cloning back, then updating all the apps and reinstalling plugins it would be just as fragmented as before.”
Here’s how the pros do that (condensed version):
- Set up a DeployStudio server
- Create .pkg installers for anything that doesn’t ship as one using JAMF Composer
- Use AutoDMG to create a custom OS X installer images to be used with DeployStudio
- Annually or biannually or quarterly create an up-to-date deployment image with updates applied
Whenever a clean install is desired, simply reboot the workstation with the N key held down to NetBoot off the DeployStudio server and re-install with the appropriate custom OS X image hosted on the server. Go to lunch and have a freshly installed and configured virgin workstation with all your apps and plugins when you get back. If you have an Open Directory server set up as well, your user folder contents can be stored independently so all your individual user preferences remain intact and get synched down the first time you log in.
Obviously, the effort to set this up and maintain it is going to be overkill for individuals and is a time commitment that only makes sense for a dedicated IT staffer at a facility, but I just wanted to share in case anyone might be interested. The only software in that list that costs any money is JAMF Composer at $99. DeployStudio and AutoDMG are free community-supported tools.
Best,
Andy -
Huh? I’m not on any betas. I’m running the GA Compressor 4.1 that released today.
/Applications/Compressor.app/Contents/PlugIns/Compressor/CompressorKit.bundle/Contents/Frameworks/Qmaster.framework/Resources/qmasterd
The stand-alone Qmaster app is no more, and the Qmaster menu in Compressor is no more, but the daemon still exists and is still part of Compressor in v4.1. The clustering functionality still seems to call upon qmasterd.Best,
Andy -
I assume you are in the loop on this?
https://forums.creativecow.net/thread/335/60147#60147
Best,
Andy -
Compressor 4.1 spawns a qmasterd process on my system and it still seems to do things when jobs are submitted. Encoding itself is handled by compressord instances in my observations.
Best,
Andy