Linked by Eugenia Loli on Thu 14th Dec 2006 08:14 UTC, submitted by teigetje
RISC OS "Gosh! I didn't realize how much discussion my original article would create. A lot of people seemed to accuse me of living in cloud cuckoo land, whereas a lot more agreed with me. I think those who disagreed have either never used RISC OS or just liked a good rant! In either case, I feel compelled to write a short follow up article clarifying some of the points I made in the original article - all of which were perfectly valid." Read the follow up article.
Thread beginning with comment 192099
To read all comments associated with this story, please click here.
Debunking graphics tests
by ectropy on Thu 14th Dec 2006 14:52 UTC
ectropy
Member since:
2006-12-14

Out of curiosity, I downloaded the source sprite file for the bar graph in Vigay's article and performed by own tests:

Using GraphicConverter 5.9.3 on OS X, I immediately 'Saved As' a 32-bit PNG file: 8,606 bytes.
Vigay's PNG generated by GraphicConverter is 9,914 bytes and dithered.

Resaving the original sprite as a GIF with GraphicConverter: 13,617 bytes.
Vigay's GIF from GraphicConverter: 14,403 bytes and dithered.

I don't know why his GraphicConverter files are larger than mine.

I then pulled my resulting PNG into Photoshop CS and used 'Save for Web' to make an optimized PNG: 3,520 bytes.
Vigay's best PNG generated by Photodesk on RiscOS: 5,480 bytes (almost 2K larger!)

Finally, I used Photoshop CS again to save an optimized GIF: 9,868 bytes.
Vigay's best GIF: 11,198 bytes (1.3K larger!)

No dithering was needed as the bar graph has only 21 colors.

I couldn't test the JPG because the link to the source file is broken.

All Paul has proven is that GraphicConverter may be inferior at optimizing graphics compared to Photodesk. Photoshop does a better job, and there are other apps available for Macs and PCs that will optimize even more. In any case, the underlying OS makes no difference.

Reply Score: 5