Apple has announced a new MacPro, is it time for the video professional to move to Windows?

hero

The wait is finally over and Apple has announced a new and more powerful MacPro, but is it really a replacement for the venerable MacPro tower that is out now. Personally I don’t think so, and this is why.

The Current MacPro is all about expandability and choice. You aren’t locked into much as you have 4 PCI slots, 4 drive bays, and 2 Optical Disc bays, so you can really expand within your machine, and can pick and chose the components you want to make the best machine for you.

The MacPro throws all that out of the window and makes a tiny little machine 9.9 inches tall and 6.6 inches wide, and only 1/8 the volume of the current MacPro is a bizarre little cylindrical tower, and it relies completely on external expandability and it’s 6 Thunderbolt 2 ports (each supposedly doing 20GB of throughput each). Inside is a single PCI 3 SSD for blazing fast speed, but all other drive space must be external, and at wity multiple monitors you will be using up at least one thunderbolt port.

For graphics it has dual AMD FirePro Workstation Class GPU’s with 6GB of VRAM each for some serious power, and can connect 3 4K displays, but you get no choice, AMD only, which is great FCP X and other programs that rely on Open GL and Open CL, but without an NVIDIA card, you get no CUDA support, which is how many top graphics programs get their insane acceleration. And most programs actually bog down with dual graphics, unless you are doing something like an NVIDIA Maximus configuration where you use a QUADRO for the graphics head and a TESLA for the rendering power, but personally I would rather go cheaper and have a new NVIDIA TITAN GTX which is basically a consumer version of the TESLA and is only a thousand, vs multiple GPU’s would should cost at least $2000 and not have CUDA support. And CUDA support rocks! Still AMD claims they are faster than CUDA.

And 20GB is fast, but not as fast as PCI. A 16x PCI bus is 64 GB of throughput, while a PCI 3 (which this Machine says it has at 40GB/s) in 16x is 128GB/s throughput. So yes you could buy a $500 expansion chassis and run a single PCI 2 4x which has a 16 GB throughput, but I would rather have that extra $500 put into my main case and be able to fit it all into one unit instead of having an octopus of cables and drives and expansion chassis connected to a new version of the Mac Cube (actually more like a Pro Mac Mini).

And then we come to the processor, which is a single XEON chip with up to an impressive 12 cores, but the real power of Xeon’s is dual processor, which could have 2 12 Core Xeon’s (though a 12 core Xeon is going to be insanely expensive, especially when you could get an 8 core [or possibly up to 12] i7 with hyperthreading for a whole lot less). If you aren’t going to have multiple processors, why not have the top of the line i7 instead, as on a mac you are not going to be overclocking.

And the rotating thing is weird, especially if all 16 connectors are full. And seriously no front USB ports? I use those all the time!

It will be a powerful machine, but it is dumbed down and not internally expandable. And will be mess if you expand everything externally. And you will have to live without CUDA, which I have grown to love with Premiere Pro and After Effects (yes Premiere will work with AMD, but so far at least CUDA has been faster).

The thing is I love Mac, but I would rather have a huge internal tower. And that means I could go Hackintosh, but they always have issues, especially with upgrades. The other option is going to Windows and getting your choice. You can get a custom build system from someone like Puget Systems (where you can get a pretty bad ass custome built dual xeon system for under $6000) or you can build your own and save money and have a more powerful machine with all the internal expansion you would ever need.

Right now I am leaning towards windows for my next machine, but it will be a while before I can afford a new system anyway, so maybe we shall see if this MacPro is good enough, but I just don’t know about giving up CUDA for a system that is more form than function!

Leave a Reply