First test.

I did all the test with 52 high resolution images.
All 52 images are in JPG with a resolution of 1920×1200, 1680×1050 etc.

Tested on a MacBook, 13 inch, 2.2 GHz Intel Core 2 Duo with 4 GB 667 MHz DDR2 SDRAM with images on the main hard disk and exporting to the main hard disk. Notice that the CoreImage-engine results can be better, but the MacBook has a ‘bad’ integrated graphical card.
Applications running in the background: MarsEdit, Safari, Mail, iChat, Finder, Intervals (a stopwatch application with the Apple Remote control support), NetNewsWire, Xcode (not compiling of course), iTunes (playing a song on the background). So the test isn’t 100% correct but it is an idea to get some numbers.

Images are exported to an existing folder, without removing the original (logical) and without removing the subfolders.

Scaling images to 997×997 pixels and using the normal export-method.

Time of export

Default engine

	First Test	Second Test
JPG	13.3 sec	14.0 sec
PNG	37.6 sec	37.0 sec
TIF	13.3 sec	13.3 sec
GIF	13.0 sec	13.0 sec

CoreImage engine

	First Test	Second Test
JPG	31.7 sec	31.5 sec
PNG	1:04.3 min	1:05.7 min
TIF	29.5 sec	30.0 sec
GIF	30.7 sec	31.3 sec

Another aspect, memory usage!

During the exports, I had an eye on the memory usage of QuickScale (measured with the top – commando in the terminal) and the real memory usage was switching between 80 – 110 MB. And that isn’t a lot if you are busy with scaling a lot of high resolution images.

The file-size

Default-Engine

CoreImage-Engine

As we can see, the file-sizes in the CoreImage are larger, but there isn’t a big difference.
This difference is a factor to the better quality of the images.
The image below shows the differences between scaling a DNG-file with the Default-Engine and the CoreImage-engine.

As conclusion.
I’m very happy with QuickScale, during the test I didn’t got any spinning wheel and I could Cancel every export whenever it is necessary. QuickScale don’t use a lot of memory and it just works very fast. The CoreImage results can better, but I think that I need to test on Mac Pro to get realistic values.

What do you think of the results?

Comments(0)