I just took delivery of a new 12 core Mac Pro.  Something about “cores” is techie right now, so I suckered for the whole enchilada   ...  2.93ghz dual hex-core.  12 cores running at that speed .. now that’s gotta be fast.

I spent most of the weekend setting it up.  I found using migration assistant would have taken about 3 days, so I opted for a rather convoluted process. First I cloned the old mac 2x1tb raid 0 to a single 2TB drive via eSata ( which only took about 9 hours). I then used it as my startup drive for my new MacPro so I could keep working, and used the old tower to setup the drives for the new MacPro. In the old MacPro I installed 4 2Tb Seagate XT Barracuda drives, each drive was partitioned with 4 segments of 50, 700, 1000  and 250 gigabytes in that order.  (This was a tip from Jack Fletcher over at GetDPI.com).  This meant the 50gb partition was on the outer rim of the platters (the fastest part of the drive), each partition relatively slower.  I then set up the first 3 partitions of all 4 into raid 0 drives, leaving me with a 200gb raid of the fastest part of each drive for the Photoshop scratch disk, a 2.8tb and 4tb  raid 0 drives for system and data, and 4 individual 250gb backup partitions.  I then cloned over my system from the single 2TB drive, and even using FW800 it only took about 4 hours.  Finally after all that was done, I moved the 4 drives over into my new MacPro and fired it up.

I launched PS and ran some tests and everything was crawling.  I ran the same tests on the old MacPro (about a 3 year old machine) and it did everything in 1/2 the time ... yes my new MacPro was twice as slow as my old one.  In activity monitor it was obvious the CPU’s weren’t even being used as it showed the CPU as about 95% idle.    It turns out all system 10.6.4’s aren’t equal, and so I took the time to re-install OS X from the DVD’s (no archive and install option anymore which I would have preferred).  Obviously something in the newer 10.6.4 enabled the new processors correctly.  I should have known better.

So the new 12core MacPro was finally faster than the previous one, but only by about 20%.  That’s not much of a gain for such a large investment and 3 year newer technology.  I started messing around with the settings in the performance preference pane, and some new options there had me puzzled.

“tall and thin”?  “Big and Fat”?  I’ll be honest, I’ve never heard of a Photoshop document described this way. Maybe this is standard jargon for Photoshop guru’s but it was meaningless to me without the little tooltip when hovering over the button. I decided to ignore those and just try a different setting.  I upped the cache to 6, and changed the cache tile size to 1024k.  After restarting Photoshop, I re-ran the tests which resulted in a dramatic speed improvement.  A test that took about 55 seconds on my old MacPro, and about 43 seconds on my new one now only took about 22 seconds on the new one.  I went back to the old MacPro to compare, but it didn’t seem to affect it that much ... actually I found had installed a plugin which did the same things as choosing the larger tile size.

Turns out this setting replaces an old tip about using a photoshop plugin called biggertiles.  In CS4 you had to find this plugin on Adobe’s site, and after installing it could see a nice performance gain in Photoshop when working with large image files.  I moved it over when installing CS5, so my old Mac was already taking advantage of this without changing the preference pane, which explains why changing the preference didn’t provide different results.  With CS5, you have control over this in the Performance preference pane.  Interestingly enough when I click Big and Flat on my new MacPro (24gb or ram) it changes the Cache Tile Size to 1024 (the number I tested), but when I select it on my MacBook Pro (8gb of ram), it stays at 128k.

As an aside, for Photoshop and probably Lightroom, I’m now suspecting the 3.33ghz 6 core version may actually be faster.  I’m not sure Photoshop and other programs really leverage all of those cores, so a faster processor speed with 6 cores might be a better choice. (there are 12 actual cores and 12 “virtual” cores, meaning when you view you CPU graph in Activity monitor, you see 24 cores of information ).  However, this assumes the only thing you do is run one of those.  As soon as you start doing multiple things at once, the cores help a lot. 

I did some tests using an Apple Developer preference panel that let me disable cores and virtual cores just to compare.  Running a Photoshop test took about 23 seconds as compared to 55 seconds on my old dual QuadCore Mac Pro.  However, turning off 8 of the cores only took a little longer ... 27 seconds. It was obvious by monitoring the cores I was only get about 30% utilization.  Then I fired up Lightroom, started a task building 1:1 previews from PhaseOne files and had Handbrake ripping a DVD at the same time.  The Photoshop task took longer but with all cores running the impact wasn’t nearly significant ... 34 seconds vs 55 seconds. All three programs churned along just fine, as though nothing much had happened. I do this quite a bit (trying to digitize my DVD library so I can stream more of my movies to AppleTV’s I have throughout my house).

And of course there is the distinct possibility that Adobe is tweaking code as we speak to improve performance by utilizing more cores.