Whither the Mac Pro?

The latest flurry of speculation in Apple circles is whether the Mac Pro has any future. This is, in essence, because Apple hasn’t revved the Mac Pro for eighteen months. Some articles I’ve read suggest this is because Apple is waiting on new Intel CPUs, but Dell’s current top-of-the-line workstations (it’s Dell, so no clue how long that link will stay good) seem to be configurable with more RAM and faster Xeons than the “current” Mac Pros. I recently started a new job and was offered a Mac Pro — I responded that I wanted to wait for Apple to rev the damn things first.

As someone pointed out, if you max out a Mac Pro’s CPUs it’s still more than twice as fast as a maxed out iMac (which is a generation ahead). (This comparison shows a 6-core Mac Pro being generally somewhat faster than a maxed out iMac.) For some value of “twice as fast”. For most of us, a nicely configured iMac with an SSD is going to be as fast or faster than a similarly priced (and thus poorly configured) Mac Pro. Oh and it will have a nice display. But it won’t be nearly as fast at “coffee break” tasks such as rendering complex 3d scenes or video comps. And you might not want the display.

What do you need a Mac Pro for anyway?

Here’s what a Mac Pro offers that you don’t get with a high-end Macbook Pro or iMac:

  • expansion slots (and large power supply to power them)
  • four easily accessible hard disk bays
  • lots of RAM capacity (64GB)
  • two slots for processor cards, allowing (for now) up to 12 cores and (in theory) easy CPU upgrades
  • extra reliability (Xeon)

Of these, the first and second can be addressed by Thunderbolt expansion (and, in fact, sticking a GPU inside a display makes a good deal of sense, except that the displays don’t exist yet, and it would be nice to be able to upgrade the GPU). I’d argue that having Thunderbolt-based hard drive housings actually makes more sense than sticking lots of hard disks in a Mac Pro.

RAM capacity is a big deal. I have 8GB in my Macbook Pro and it’s not enough. And I’m not doing anything nearly as demanding as a lot of Mac Pro users. If I could have 16GB I would. It’s easy to imagine needing more.

Upgradable CPUs are totally awesome. My favorite pre-Steve’s-second-coming Macs were the “sidecar” Power Mac 7300, 7500, and 7600, all of which had CPU daughter cards. Two years after I bought my 7600 I upgraded it with the latest PowerPC 604e for very little money and it took less than five minutes. Even if you could get a laptop or iMac with an upgradeable CPU (I had a Powerbook 1400, and upgraded its CPU too) the chances are its other limitations (RAM for example) will be so crippling that a new CPU won’t be much use (my PB1400’s hard disk and RAM largely negated the significant CPU upgrade). Apple’s professional Macs, by contrast, usually have far more RAM capacity than most pros need or can afford to fill.

Upgradeability aside, in terms of raw CPU performance I’m not sure the Mac Pro really justifies its existence any more. If you can distribute a task among 12 cores efficiently, chances are you can distribute it among a bunch of small machines (Mac Mini servers, for example) efficiently as well. In most cases, you get more bang for your buck from Mac Mini servers (with quad-core i7s and 8GB of RAM each) than from Mac Pros. And generally this is a net win for the user (e.g. if my storage is networked and I can hand off a task to a bunch of network nodes, the computer I’m actually using is going to be a lot less bogged down), and there’s also the fact that another computer is another computer. It’s easy to imagine having a Mac Mini hooked up to each TV in the house and also acting as a small render farm. For 3d rendering, there’s simply no way that the Mac Pro is cost effective — go to the cloud for short-term needs*, and build a render farm of whatever floats your boat for long-term needs.

Note: * the going rate for cloud-based render farms seems to be $0.70 per core per hour ($0.50 in bulk). Setting aside the fact that they’ve set up nice interfaces for kicking off batched renders and the like, that works out to a Mac Mini Server (maxed out with RAM, at retail) paying for itself in under three weeks. So if you want to run your own rendering farm, the Mac Mini Server is cost effective both initially and at scale.

As for the Xeon reliability advantage. I guess you know if you need it.

The Mac Pro Use Case

Rendering: I’m not doing any 3d or video production right now. If I were rendering long 3d sequences and video compositions, I would definitely not want to do them on my Macbook Pro, but I wouldn’t do them on Mac Pros either. Buying a bunch of Mac Mini servers to handle rendering tasks is not only more cost-effective than Mac Pros, but more granular and scalable. If I need 120 cores to render a 3d short, I can buy 30 Mac Minis at $36,000 or  10 Mac Pros for $60500 (or, better yet, rent the CPUs in the cloud for a whole lot less). Coming back down to Earth, I can buy one Mac Mini for $1200 to offload my rendering tasks and completely free up my Macbook Pro. I’d rather render stuff on a separate computer slowly than quickly on the computer I’m trying to use for other stuff.

Interactivity: The ultimate use case for any high-end workstation comes when you want interactivity. It’s not for rendering a movie, but rendering one frame so you can make adjustments. Or being able to perform some operation interactively at 10fps vs 3fps. Here, there’s no question that the Mac Pro has no replacement now or in the foreseeable future. If I want to mess around with billion polygon scenes or hundreds of layers or effects nodes interactively, chances are I want the most CPU, GPU, and RAM I can possibly throw at the problem. And chances are I don’t care if it’s portable.

I suspect that most high-end 3d and video guys can live without a Mac Pro simply by working smarter. I suspect a lot of them use laptops most of the time and work around bottlenecks. What’s more, a lot of them are probably using old hardware because it’s “good enough” and while they’d love to have more of everything, they don’t really need it.

Licensing Mac OS X

Another option would be for Apple to license Mac OS X in a restricted way for third-party hardware. (It’s already happening in a way — 10.7 is able to run virtualized.) Years ago many of us discussed the possibility of Mac OS X being licensed for “clones” and one persistent option was “once Apple didn’t care about Mac revenue”. The problem with licensing Mac OS X for the high end market would be that one could easily see it cannibalizing the high-end consumer market. Even if a “Mac OS X for clones” license costs $1000 it’s going to be quite attractive relative to paying $3000 for a maxed out 27″ iMac.

It’s also possible that Apple will simply continue to turn a blind eye to clones. A lot of 3d artists I know simply build themselves a Mac-compatible clone. (You can build a pretty good Mac Pro substitute for under $1500.)

It’s probably easier for Apple to cater to Mac Pro users simply by making the minimal, obvious improvements to the Mac Pro line or, if that becomes too onerous, simply selling generic compatible clones using off-the-shelf parts (i.e. what Dell does) in a nice box. This would be simpler than coming up with complicated licensing schemes and address a real need.

Gaming

I’m no longer a hardcore gamer. (Parenthood, etc.) My gaming needs are largely satisfied by my iOS devices and my XBox 360. But there is a hardcore gamer market out there and some of them (exemplified by John Siracusa) use Macs. There’s no question that a Mac Pro is a wonderful games machine (and that’s actually how I justified my purchase of a Mac Pro back in 2006 — since I used to buy PCs to play games on and Macs to work on, I figured it would be a net saving over a less expensive Mac and a PC for gaming). But I see a lot of hardcore gamers using laptops these days simply because they’re more convenient.

You could try to make the case that if Apple were to ditch the Mac Pro, it could finally release the mythical xMac to cater for gamers, but I suspect that even if it were priced compellingly (say $1000), most of its target audience would think long and hard about it, and then get a Macbook Pro or iMac anyway. Still, it’s pretty weird that Apple doesn’t have a Mac Mini configuration with a decent GPU and quad core CPU which would be 90% of an xMac right there.

Conclusion

In terms of computing requirements, my needs are probably greater than 99% of other users, but not 99.9%, and, guess what: I don’t really have any use for the Mac Pro any more. Does this mean that the target market for the Mac Pro is somewhere between 1% and 0.1% of the PC market? Of the five points I mentioned at the outset, the only one I really take advantage of is the hard disk bays, and a Mac Mini with an external enclosure would do as well. Better actually, since it would use less room and less power.

Indeed, all of the Mac Pro’s advantages are offset by its bulk. For me, the difference between having that extra RAM and grunt is offset by being able to take my “digital life” with me in a backpack anywhere I go without worrying which files are where. I would be far happier with the ability to dock with extra horsepower and storage. Indeed, if Apple moves towards allowing a single thunderbolt connector to add a whole bunch of extra oomph to your Macbook Pro or Mac Mini this may not only serve the needs of the kinds of people who use Mac Pros but new kinds of power user (e.g. the family with huge amounts of digital media that doesn’t want to invest in a Mac Pro).

I’d love to see another generation of Mac Pros, but I won’t be too sad if I don’t, and I doubt anyone will care in 2013.