It doesn't seem that useful a test to me - it starts with a fairly large image, resamples it to 300%, changes it to CMYK, Lab then back to RGB, then creates something like 9 layers. That's a far bigger file than most of us work with, and they're not typical workflow steps.
Beware, too, that the action changes your colour settings.
On my machine, I managed to get 3:18 (Photoshop CS2 9.0.1, Dual Xeon 2.66GHz, Windows XP Professional SP2 with the /3GB switch, 3GB of RAM, one 15krpm U320 SCSI 36GB hard disk with the OS and program files, one 15krpm U320 SCSI 36GB hard disk with my swap file, one 7200rpm SATA-300 500GB hard disk that had scratch space on it).
For the test I allowed Photoshop to use 2.5GB of RAM (typical for me - I have about the only setup on which Photoshop can use more than 2GB of RAM) and set Cache Levels to 1 (normally I have them at 4). I didn't hide my palettes - that may make a difference.
Watching what was going on with the Windows Task Manager shows that Windows starts doing rather odd things with file caching and swap file usage. Initially all stays in RAM, and the VM then RAM usage shoots up to the maximum. When the Shadow / Highlight step is hit, CPU usage drops to next to nothing (up until now, some of the time it's 50% indicating complete use of one of my two processors, often it's 90ish%, indicating both processors are being used) and the RAM usage drops way down.
It's at this point that scratch file usage is up to over 5GB (normally I keep my scratch space on the same SCSI disk as the program, and that disk ran out of space the first time I tried it), and processor usage doesn't recover until a lot of disk I/O has happened. This means that things like my antivirus software are in play, also Windows doesn't necessarily behave that sensibly with regard to memory allocation when both program memory and I/O are both high (Windows often inexplicably pages out program memory in this scenario, whereas UNIX kernels seem to manage better in this sort of scenario, so Mac results may well better Windows).
As I said, I don't regard it as a good benchmark. In real post-processing use, I never have 2.5GB of RAM and 5GB of scratch space in use when working on a single file - indeed, since moving to 3GB of RAM, setting the /3GB flag in boot.ini (without it, no program can use more than 2GB of RAM - you do need to be using Windows XP Professional SP2 for this to work properly) and allowing Photoshop to use up to 2.5GB of my 3GB RAM, I've only rarely seen Efficiency drop below 100%, indicating that Photoshop is working entirely in RAM. The only thing this did show is that I've got my settings right to allow Photoshop to use more than 2GB of RAM!
Applying the 9.0.1 patch does cut down some delays, especially some of the "temporary hangs" I'd grown used to.
David
