AJBMatrix said:
True about the Photoshop except one thing. The real-time previews of what you are changing. With larger filters and larger files you may notice it being slower with less VRAM. I have been there. There is a difference on older, slower RAM machines between 128, and 256. Photoshop does use Video RAM but not much until doing these previews and changing settings.
It was my impression that Photoshop doesn't do any 'native' useage of VRAM. i.e. It only uses VRAM in that the OS loads the Photoshop window into VRAM.
I can't find any definitive answer on Adobe's website, and my personal experience is with pre-CS versions; so it's entirely possible that CS and/or CS2 use it. Basically, if they use CoreImage, then they can more easily take advantage of VRAM, but I can't find any info that even CS2 uses CoreImage.
All I could find is an anecdotal report that moving form a Matrox 550 with 32MB VRAM to an ATI 9250 with 128 MB VRAM made a huge improvement in Windows. That could be easily attributed to the more modern video chip having better Windows driver support than from any inherent Photoshop improvement.
For example, on Windows (and pre-Quartz Extreme Mac OS) systems, things are only loaded into VRAM
as they look on the final monitor. Meaning if you have a 4000x6000 pixel image open in Photoshop, on a 1024x768 screen, only the 1024x768 part of it is loaded into VRAM. With Quartz Extreme, every separate window is in VRAM, meaning if you have 10 800x600 windows open in 24-bit color mode, you're using 14.4 MB of VRAM for those windows, plus any for the desktop background. But even with QE, it still doesn't load the entire 4000x6000 image into VRAM. With CoreImage, and a properly written application, it COULD load the entire 4000x6000 image into VRAM, but even that large an image would only be 72 MB. But, again, I can't find any info that Photoshop uses CI. (I would appreciate any info pointing me to the fact that it does, though.)
If Photoshop
does use CoreImage, this would mean that it should run at near-100% speeds through Rosetta, because it means that most of the app isn't really emulated. (Anything that uses CoreX, or OpenGL, won't need to be emulated, because those parts were written in processor-independent code in the first place.) Filters are still emulated, though. So would still suffer from emulation lag.