We closed this forum 18 June 2010. It has served us well since 2005 as the ALPHA forum did before it from 2002 to 2005. New discussions are ongoing at the new URL http://forum.processing.org. You'll need to sign up and get a new user account. We're sorry about that inconvenience, but we think it's better in the long run. The content on this forum will remain online.
Page Index Toggle Pages: 1
OpenGL AA Tests (Read 1309 times)
OpenGL AA Tests
Apr 13th, 2007, 6:46pm
 
In short, am wondering if people would like to participate in a quality assessment program.

Long Version...

Situation:  I'm in the market for a new video card, and I'm more concerned with quality than performance. (not that killer performance wouldn't be appreciated, but only if the quality is there first)

Premise:  Antialiasing quality varies A LOT among video cards.  Specifically as regards raw geometry rendering (strokes and fills) moreso than texturing.

Resources:  Most video benchmarks out there only focus on polygon counts and fill rates and other performance stuff, with little said to quality, except perhaps to say "..such-and-such 3d game textures look amazing with 16X anisotropic filtering..".  But if you then look at the raw lines drawn on one of these much-raved-about cards you may be suitably unimpressed.  (despite those amazing texturing abilities)  It would seem that "how it looks in a Processing sketch" is not a major concern of video card manufacturers - can you believe it?!  Cheesy

Research:  I'm currently fortunate (or unfortunate) enough to have easy access to five different video cards.  So I wrote up a quick testing sketch and grabbed output from each to compare.  Guess what?  Vast differences in quality.  No surprise there, I already knew, just lacked the hard evidence.  Further, one card that claims up to 16X FSAA produced bit-for-bit identical results at 4X, 8X and 16X! (no improvement beyond 4X - and the 4X is pretty poor)

Dilemna:  I'd like to see output from an even wider variety of cards, so that I could make a better purchasing decision, but can hardly afford to buy 40 cards in the hope of finding one good one.

Idea:  So... am wondering if others out there in the Processing community would like to pool resources and run a similar test on their own cards.  Then we could post results to some shared location (a flickr group?) for all to benefit from.  You could run my test, or one of your own design, or some commercial benchmark, whatever - as long as the output image demonstrated likely artifacts you'd encounter from a sketch rather than from Quake.

Any interest?  Or know of such results that already exist?

Cheers,
Dave
Re: OpenGL AA Tests
Reply #1 - Apr 14th, 2007, 5:56pm
 
Sounds like a great idea. I only have access to Macs but I would be very interested in seeing the variation between cards. A flickr group would work well for this. Let me know if I can help.

stephen
Re: OpenGL AA Tests
Reply #2 - Apr 14th, 2007, 10:49pm
 
Hi Stephen, glad to have you aboard.  Macs are fine - I'd be interested in results from any platform, since they'd still provide useful measures of what's out there.  So I've started what will surely become the most popular group ever conceived on flickr:  Wink

http://www.flickr.com/groups/openglforartists/

A couple results posted there, and sample code posted in the discussion.  All submissions welcomed, the more the merrier!  I figure it's an experiment, give it some time, see if any participation occurs.
Re: OpenGL AA Tests
Reply #3 - Apr 23rd, 2007, 5:24am
 
I was getting ok results with my quadro 3000- the 16x was noticeably better than the 4x, but still not as nice as an image rendered larger with a bicubic or bilinear scaling.
Re: OpenGL AA Tests
Reply #4 - Apr 23rd, 2007, 8:59pm
 
I ran your test program also, and it was very helpful in seeing the differences in AA settings.  My card only allows 2x,4x and 6x.  But the difference was still very noticable.

Once I ran the program to generate the AA test images, I put each one into a layer in a Photoshop document.  By doing this, it was very easy to A/B the different layers and compare the results.

Re: OpenGL AA Tests
Reply #5 - Oct 8th, 2007, 12:06pm
 
I ran your test program and posted them on the group
I was using a NVIDIA Quadro FX 540.
Page Index Toggle Pages: 1