Monday, January 12, 2009

Core Image benchmark comparison (Geforce 7600GT, Geforce 9600GT and Radeon 3850)

I use my Macs and OSX86 machines also for photo editing with Aperture. The Apple Pro-Apps (Aperture, Finalcut, etc.) rely on Core Image and therefore profit heavily from a fast graphics card in the system.

To compare different models of systems and graphics cards, I made some comparison benchmarking tests using iMaginator 4 (trial version).

The test consists of exporting a morphing movie of 1000 frames, which contains all transition effects in the order they are listed in the selection pane (see below) using the two default images (leopard and rose) as source and destination for the morphing.

These were the "Transisions" used (top to down):

And I changed the following preferences (only changed # Frames to 1000):

Then I saved the movie using "Save to Movie" from the "File" menu, started the stop watch when I clicked the "Save" button, moved the mouse, so the cursor went to the "Beachball". Then I waited until the cursor turned back to the normal pointer and stopped the stop watch.

These were the results on Mac OS 10.5.5 (faster = better):

iMac G5 2.0 GHz ATI X800: 6m50s
iMac C2D 2.33 GHz Geforce 7600 GT: 2m14s
OSX86 C2Q 2.66 GHz ATI Radeon 3850: 1m22s
OSX86 C2Q 2.83 GHz Geforce 9600GT: 1m21s

I was surpised to see the Geforce 9600GT perform that fast, although that configuration is slightly faster and has more cache for the CPU than the one with the Radeon 3850. But I guess it is safe to say that the Geforce 9600GT is not such a bad choice for Core Image intensive apps. I think a Radeon 3870 would probably out-perform the Geforce 9600GT. Maybe a Geforce 9800GT could be faster than a Radeon 3870, but unfortunately I can't test that since I do not have these graphics cards at hand.

In a later post I might add more details about the system configurations and maybe will also put some tests done regarding OpenGL performance of the same systems (where the Geforce 9600GT also scored  better than the ATI Radeon 3850).

No comments: