PDA

View Full Version : How do I benchmark a machine with Photoshop?



Mike Lewis
9-Nov-2006, 18:05
For those of you who are using Photoshop to edit your large LF image files: suppose you had the opportunity to use Photoshop on new, fast hardware to edit a large (400MB+) image file. If you wanted to time certain Photoshop actions on the file, which ones would you try?

Ed Richards
9-Nov-2006, 18:19
I would use the actions function to develop a series of time consuming things to do, then run it on a directory with several files, even if they are just copies of the same file. I would resize, say down by 50% then back to the original size, then add a couple of histogram lawyers, then use smart sharpen (not that I would ever use this in the real world, but it does waste time), then save it out. What you want is something that takes enough time to get a good measure when it is run across several files. One problem - if you run out memory, all the times explode.

Frank Petronio
9-Nov-2006, 19:10
Disk intensive operations like saving don't really depend on the processor speed, at least if you are comparing machines from generation to generation... operations that require crunching lots of numbers -- distortions and transformations -- would be a good to include, along with sharpening, converting color modes, etc.

It is really hard to benchmark the painting tools. In the real world, you just work on the same file using both machines and see if you can sense the difference. If I don't see a huge difference -- like 2X -- I wouldn't bother upgrading.

Gaining 50% better performance doesn't reduce your production time that much, depending on how thoughtful (or indecisive) you are while working.

I think upgrading with processor generations make sense... G3, G4, G5, Core 2, etc.

In a busy digital lab with expert users doing repetitive tasks 24/7 I think the subtle speed bumps would be more important. I'm talking about an individual photographer working 10-20 hours per week in PS.

Henry Ambrose
9-Nov-2006, 19:56
I probably wouldn't try to develop my own benchmarks unless what I did was very specialized and repetitive, day in day out. What I would do is use something like the retouchartists set of actions -- http://retouchartists.com/pages/speedtest.html

This will let you compare your stuff to configurations you may not have right at hand.

To save you some time and work -- in general, for PS work, Quad G5s are fastest followed by Dual G5s then the new MacPro. That will change in the spring when Adobe releases the Universal Binary version of Photoshop. Then the Intel Macs like the MacPro will be the kings. And as always, all the RAM you can afford and install and the machine can address is the best thing to have for PS.

A 400 MB file is not that much to push around with anything from the early G4 towers onward, unless its an everyday, all day ritual. If you have plenty of RAM (maxed for whatever machine you have) you won't be hitting the scratch disk. Unless you do lots of layers then it gets out of hand. But layers are for sissys anyway (that oughta get something started - grin)

Once you move outside of what can happen IN RAM on your machine then you have to get going on super fast scratch disks, fiber channel, SATA RAIDS, etc. Which I'd consider a waste of money unless you have the fastest machine already and its the last ditch effort to get performance. If you have an older machine thats not so fast you'd be better off spending all that loot on a new box. But again, at 400MB you're not really plowing that big a furrow. Still painful to watch that progress bar though.

What Frank wrote about speed is important unless you have lots of cash to spend. (and if you do then good on you) The latest hyped super fastest thing ever for someone who spends 20 hours a week in PS is kind of a luxury in that it won't really save you that much time. It needs to be LOTS faster than what you have now to really make a difference at the end of eight hours solid PSing.