Adobe Photoshop CS3 Performance

Highly Efficient Rosetta or Darn Slow Photoshop?

When Apple introduced Intel-based Macs, a year ago, it provided comprehensive migration support, as well. Free compiler (program "translating" human-readable source code of programs into "machine code") shipping with OS X is able to produce so-called Universal Binary - code that both old and new processors can understand. So, for many software vendors migration was just a matter of recompiling their source code. Unfortunately, it was more complicated and time-consuming for vendors of large systems where recompilation is not as trivial. Several large software vendors were not able to have their code recompiled and debugged, yet.

To support those more inert, Apple developed technology called Rosetta. Rosetta is a virtualization engine that allows programs, written for old, G# processors to run on Intel-based Macs without changing single line of code. Microsoft Office for Mac, still runs on top of Rosetta, with no plans to recompile it into Universal Binary. Adobe, also, did not have universal binary version of its graphical applications, until recently. In Adobe's case they were very eager to complete the migration, though, as virtualization is a performance hit and performance is important for a graphical information processing software like that from Adobe.

The highlight of and the biggest advancement in the recently released Adobe Creative Suite 3 is the fact that it is compiled in universal binary - with full support of Intel processors and avoiding Rosetta. Adobe is very excited about it and has been honking all over the press how much better/faster the new version is.

Is it really?

A friend of mine got CS3 and we did a quick-n-dirty test on it. CS3 does start-up pretty quickly compared to the under-Rosetta CS2 , but that's about it. We had both versions apply a 100%, best-quality radial spin blur to a hi-rez (3004x3760, 1.1MB) photo on a Macbook Core Duo with 1.5GB RAM. Results are as follows:

CS2 (w/ Rosetta): 2.04 sec
CS3: 1.33 sec

That, my friends is just 53% increase in speed and, honestly, would hardly impress anybody, even in a quick-and-dirty test. Either Rosetta is incredibly efficient (kudos to Apple) or Adobe did not do too good of a job, or both.

Concluding analysis, we need to mention another metrics, as well, to be fair. The memory footprint of CS3 was significantly lower - 80MB vs 255MB consumed by CS2. Not bad.

Not the best test?

Hi - I've been looking for anything on the web that details how much faster CS3 is when running Natively as opposed to in Rosetta on an Intel iMac, the reason being that I have a couple of filters (that I barely use) that only run in Rosetta at the moment, and I'm not sure I want to run the whole app in Rosetta just for them.

I have to say, a 53% speed increase natively might not mean much in a 2-second process, but if you've got an hour's worth of batch processing to do, that's a whole 30 minutes extra under Rosetta. Secondly, it wasn't exactly a deep test. Why not try running a much more complex action and see how it holds up under the strain? I think I might try this myself.

Sure

Oh, sure. Not only it was not "the best" test, it was not even close to comprehensive. Just a quick-n-dirty one to get a feel.

It would be great to hear about the results of your tests, when you get a chance to do some.

cheers