Snow Leopard Test Notes
To test Snow Leopard (Mac OS X 10.6) compared to Leopard (Mac OS X 10.5.8), I created two partitions on an external Firewire 800 drive, each partition being 80GB. I cloned my Leopard boot volume to both partitions, then updated the one of them to Snow Leopard.
This gave me two bootable partitions on the same drive with nearly identical performance characteristics. In less than a minute I could reboot into 10.5 or 10.6 (32-bit or 64-bit), thus allowing comparison of specific features on exactly the same hardware, with exactly the same apps, preferences, etc. There can be no more fair comparison than that.
Hardware appropriate to the task. I did not, for example, test Photoshop CS4 with 3 or 6GB of memory, a self-defeating configuration for a Mac Pro (though many tasks will do just fine with 6GB). If your time is worth something, get memory and drives which eliminate any bottlenecks. Looking for specific advice? Contact me for a consultation.
Unless otherwise noted, test machine was a quad-core 2.66GHz Mac Pro Nehalem, with 12GB of memory, and a 4-drive striped RAID for data, ensuring adequate disk speed to rule that out as a factor in tests involving Photoshop CS4 and similar programs.
CPU-intensive tests that would benefit from an 8-core machine used the dual 2.93GHz Mac Pro with 24GB.