Mac Pro Users: Would you need 12 cores?

[Brad Owens]

With the announcement of the new Mac Pro models coming Apple - Mac Pro - Now up to 12 cores of processing power. I'm starting to save my pennies but having trouble deciding between options

My question is, with hyper threading, the new 6 core processor is capable of running 12 cores, that would mean the 12 core processor can run 24 CORES! Can you see a reason I'd need the 12 core or would I be perfectly fine with the 6 core?

I'm looking to run Adobe Lightroom and the Final Cut Suite. Assuming Apple upgrades their Final Cut Suite this year to handle multiple cores, is it worth it for me to hold off on a new camera and put my money into the 12 core processor? It goes without saying that I'm going to throw as much RAM in there as I can afford and run a SSD as the boot disk with the other three drives set up RAID 0. I get lost when it comes to processors though.

I'm currently running an early 2008 MBP with a 2.6 dual core and 4gb of memory so any upgrade will be faster than what I have. With being in the market for a new camera to upgrade my HV30, I just can't see getting a higher bitrate camera with a less than perfect machine to edit the footage on so a computer may be my first purchase.

[Andy Wilkinson]

Good questions and reasoning - and similar to the ones I had a year ago when I bought my 8 core 2.66Ghz Nehalem Mac Pro with 12GB RAM and Four 1TB 7,200 RPM hard drives, i.e. to future proof myself as much as possible "core wise" for when we get "a real FCS upgrade".

So one year later I have a lovely MP and do lots of commercial work on it/still very pleased with it's performance and stability - but rarely do I use it's true potential because of the current software limitations with FCS2. Sure it's faster than my 2.5Ghz 2008 15-inch MBP with 4GB RAM 7,200 RPM 200GB drive (i.e. similar spec to your MBP), but not dramatically faster with simple or moderately complex video editing (except in rendering in Compressor etc.) when the MBP is hooked up to one of several G-RAID 2TB drives via FW800 (mostly XDCAM EX3 workflow). Obviously, when things get serious the MP is king.

I paid a lot of money (over what I needed to) to have this future proofing/flexibility with this particular MP. Only you can decide if that's a path you want or need to go down (or can afford!) but I sure hope we see that long awaited (true) upgrade to FCS soon! I'm hoping this MP will serve my business needs well for the next 3-5 years so took a long term strategic view and bought the best I could afford at the time - a path that has often served me well (once the short term pain of paying for it all has gone!). And if FCS does not get upgraded to use all those cores soon I'll switch to another NLE - they are just tools to get a job done after all.

[William Hohauser]

Excellent new Mac, but on the FCS front it might make sense to wait (especially in the production recession many are experiencing right now) to see if the long rumored 64 bit FCP 8 will improve the performance of all MacPros. My experience has been that significant speed improvements come with Final Cut updates. Right now Final Cut recognizes multiple cores but not multiple processors. Render some effects and watch the activity monitor utility if you have dual processors in your Mac. It would be great for my work just for FCP to use all eight cores when I have to render some intensive filters.

[Robert Lane]

Williams' advice is spot-on; neither FCP6 nor 7 can fully utilize the hardware currently available so making an upgrade purchase would not be wise - yet. We're all praying for FCP 8 to be released this year (well, asap would be more accurate) but until that happens and we see what it will and won't handle then it's time to wait and hold onto your hard-earned cash.

[Dave Partington]

I currently have 8 cores and 16GB ram. I could always use more cores - especially when compressor is running! On large jobs I'm sending stuff to other computers in my render farm and still wish for more!

[Robert Lane]

Unfortunately the way computing mathematics work out, even if you had a newer 12-core system - now available for ordering - you would not see a major speed boost in Compressor or other CPU-heavy apps. Not until they become 64-bit and can fully address all the dual-processors and RAM available on the system.

Anyone who currently owns a MacPro would be wise to wait. If you don't own a Mac tower at all and need one then order the best system you can afford and be happy.

[Chris Korrow]

I think the most important question is - How much are you editing? If your a full time editor, or if your someone like me, that does everything and maybe 30% of my time is spent in FC.

I've got a 2.93 quad & am very happy with the speed & stability. Way faster than my MBP. If your dealing with getting stuff out the door ASAP, then the 12 cores can help, But if you just need to render something overnight & & burn a DVD the next morning every few weeks, I'd think the 6 core is plenty.

[Craig Seeman]

There are a few things that make me wary of this MacPro update. Apple's treating it as minor . . . a speed bump with more cores. Missing is new ports either and/or USB3 and LightPeak. Also people are looking at nVidia graphic cards for FCP to have its equivalent to Premiere's Mercury Engine. I have a hunch we'll be seeing another update in a few months probably around the time of FCS4. My guess that will happen between December and March.

BTW the monitor situation is getting odd too. 24" and 30" going away and being replaced by a glossy 27" Locking people into one Apple monitor, glossy no less, is not ideal for an editor IMHO.

All in all it looks very "transitional" at the moment. I'd really wait for LightPeak if you already have a MacPro. If you're buying one now for the first time, in 6 months you might need to use PCIe slot for the LightPeak adaptor when the time comes.

[Dave Partington]

Originally Posted by Robert Lane

Unfortunately the way computing mathematics work out, even if you had a newer 12-core system - now available for ordering - you would not see a major speed boost in Compressor or other CPU-heavy apps.

Not quite true (and yes - I was a programmer for 25 years - so I know a little of what I am talking about).

If you use compressor as a standalone app then you would be right. But, if you have a multi-core system you should set up QMaster - even on your local machine and use it as a local render farm. This can then make use of ALL the cores and it significantly improves speed / decreases render time. The more cores you have the better it gets.

For FCP - I would not expect to see much difference.

[Brad Owens]

These are all incredibly helpful.

Since I'm just starting into event/corporate video, I'm guessing that the top of the line (read: out of my price range) may not be the best move. Lightroom will run fine on either of them and FCS will be fast "enough" with 6-8 cores. No use in paying the premium for 12. Maybe in two years when I'm cranking out business, then the time to upgrade to the bigger, badder machine makes sense.

Thanks everyone. Saved me some cash and some headaches.

[Robert Lane]

Originally Posted by Dave Partington

Not quite true (and yes - I was a programmer for 25 years - so I know a little of what I am talking about).

If you use compressor as a standalone app then you would be right. But, if you have a multi-core system you should set up QMaster - even on your local machine and use it as a local render farm. This can then make use of ALL the cores and it significantly improves speed / decreases render time. The more cores you have the better it gets.

For FCP - I would not expect to see much difference.

That would be great suggestion except for a few niggles with Q-Master clusters:

1. The majority of FCP users even power users have no clue how to properly setup a virtual cluster properly and end up getting Compressor wonky in the process. It's not an intuitive process and it even took me a few days to get it right years ago. (see Brian Gary's Compressor Quick Reference Guide for a good explanation of how to do this - if you dare)

2. A virtual cluster only affects and speeds-up MPEG2/4 encodes because it is multi-thread capable. H.264 and most other codecs are not and are therefore totally unaffected by a virtual cluster's availability, hence no speed increase. That's why all the hardware-based H.264 accelerators have come to market because there is currently no software available that can force multi-threading in H.264.

So if all someone did was MPEG encodes for DVD then yes, a virtual cluster would be of great benefit on any multi-core system. But considering H.264 has outpaced MPEG usage by a wide margin then my original suggestion stands: An upgrade to a 12-core system - at this time - is unwise, unwarranted.

[Craig Seeman]

I actually think it's very easy to setup a virtual cluster.
QMaster prefpane
Check QuickCluster with services
check boxes next to Compressor Distributed processing for Compressor
Options for selected service - select the number of instances.
In Compressor select the QuickCluster name instead of "this computer."

H.264 is MPEG4 Part 10 and certainly takes advantage of QuickCluster.
I wrote a review of Matrox CompressHD "elsewhere" in which I noted the speed difference between no cluster, 8 Instance QuickCluster, CompressHD. CompressHD was many times faster than the QuickCluster but QuickCluster was certainly faster than no cluster.

[William Hohauser]

My h.264 renders are faster with QuickCluster enabled and I monitor the cores to see this in effect.

Setting up QuickCluster is not easy in the Mac sense. It's more like working with a niggling WindowsXP system to get it right. The instructions on the web are somewhat contradictory and it took a few days of trial and error to get it right. My QuickCluster set-up is not exactly according to any of the instructions I found. On my computer four instances caused severe CPU lockups when working with reference files but not self contained files. By dropping down to 2 instances I attained the same high speed with all files. This might be due to the amount of memory I have, 6gb. I believe ProRes renders in Compressor are multi-threaded as well. I'll do some in a few weeks. 12 cores under QuickCluster would certainly speed up DVD and h.264 rendering in Compressor. An hour SD program using default setting 2-pass MPEG rendering should easily come in under 15 minutes with 12 cores.

[Dave Partington]

Originally Posted by Robert Lane

That would be great suggestion except for a few niggles with Q-Master clusters:

1. The majority of FCP users even power users have no clue how to properly setup a virtual cluster properly and end up getting Compressor wonky in the process. It's not an intuitive process and it even took me a few days to get it right years ago. (see Brian Gary's Compressor Quick Reference Guide for a good explanation of how to do this - if you dare)

2. A virtual cluster only affects and speeds-up MPEG2/4 encodes because it is multi-thread capable. H.264 and most other codecs are not and are therefore totally unaffected by a virtual cluster's availability, hence no speed increase. That's why all the hardware-based H.264 accelerators have come to market because there is currently no software available that can force multi-threading in H.264.

So if all someone did was MPEG encodes for DVD then yes, a virtual cluster would be of great benefit on any multi-core system. But considering H.264 has outpaced MPEG usage by a wide margin then my original suggestion stands: An upgrade to a 12-core system - at this time - is unwise, unwarranted.

Setting up a cluster is TRIVIAL and it works for ALL codecs as far as I can tell (and yes I use quite a few - including H.264).

If people are struggling it wouldn't take too long to make a quick tutorial video...

Compressing using multicores in a quick cluster works by chopping up segments of the video and processing them individually by each core - so a 12 minute video could get chopped up in to sections something like this (not guaranteed to be exactly like this)

2 x 6 minute sections for a 2 core
4 x 3 minute sections for a 4 core
8 x1.5 minute sections for an 8 core
12 x 1 minute sections for a 12 core.
Audio is also processed separately - but as a complete 12 minute section.

You didn't need multithreading to do this.

There is an overhead incurred by splitting the file and then putting it back together again, but for anything other than a tiny file it's usually worth it.

[Robert Lane]

Originally Posted by Dave Partington

Setting up a cluster is TRIVIAL...

Super, except the vast majority of readers of DVinfo and the people I've taught and sat next to at various workshops don't share your "trivial" attitude towards a very details-oriented task. Virtual clusters makes sense...once you understand the proper setup but getting that understanding isn't intuitive and not as easy nor as trivial as you make it out to be.

Can the average user do it? Of course. Is it as easy as as setting up a timeline? No. My point being that the devil is in the details with VC's and FCP 6 users are the ones with the greatest issue with this setup.