Hi Eric,

You are correct that what I saw was much like your Artifacts128.png.

Most of my stress testing has been on ATI since its artifacts are so obvious; I
haven't tried hard to cause on NV but I'll try when I have a chance.

I'll try to adjust the parameter and report my results.
Where exactly is the offset parameter to adjust? (sorry I'm a bit pressed for 
time)

Thanks
-- mew




E. Wing wrote:
>> Date: Tue, 11 Jul 2006 10:09:32 +0100
>> From: "Robert Osfield" <[EMAIL PROTECTED]>
>> Subject: Re: [osg-users] sigh: osglogo 1.1rc1 on ATI
> 
> 
>> Hi Eric/Mike,
>>
>> I think for 1.1 we should experiment with what mulitpliers work on
>> different hardware, and then come up with some auto detection code
>> that detects the hardware and produces an appropriate fudge setting
>> for it.  If could be a simple as its ATI then use 512 as a multiplier,
>> it could be that this should go into osg::PolygonOffset itself.
> 
> 
> Unfortunately, I don't think there is one perfect value for the ATI
> cards. I've been experimenting with the values more and there seem to
> be two different z-fighting cases that conflict with each other. For
> small multipliers, I'm getting the artifacts that I originally
> complained about and I think Mike is also complaining about. (see
> attached Artifacts128.png) You see holes/gaps in the the glyphs and if
> you rotate the text quickly, they come and go in different spots very
> quickly which gives an unplesant  'flickering' effect. Pumping up the
> fudge multiplier helps remove this effect. I noticed that Nvidia gets
> it too occasionally at the current 2x multiplier. Pumping this value
> up helped there as well. For ATI, I found that 128 still has some of
> these artifacts (which the picture is set to). Going up to 150 might
> be the minimal acceptable value where flickers are few, though I think
> I'm slightly more comfortable with 192. At 512, I don't notice these
> effects any more. (FYI, if I disable the textures, and just look at
> the polygons, there are no holes/artifacts so its partially related to
> texturing.)
> 
> But pumping up the offset multiplier too high seems to lead to a
> different z-fighting problem. (see attached Missing192.png). When
> zoomed in, and looking at extreme angles, it seems that some or all of
> the backdrop polygons completely disappear from being rendered in the
> distance. Higher the multiplier values seem to cause more backdrop
> text to disappear. (Nvidia doesn't seem to suffer from this value.)
> (And FYI, if I disable the textures, the polygons still disappear on
> the ATI, so unlike the first case, this does not seem related to
> texturing.)
> 
> So unfortunately, there doesn't seem to be a perfect value between the
> artifacts and the missing polygons. After experimenting, I'm thinking,
> I'm thinking 192 might be the best balance between the two. But I
> think the range I'm looking at is 150-512.
> 
> I am curious to hear what Mike sees for different values.
> 
> But I'm thinking we might have to expose this value to a public API.
> 
> Btw, my ATI card is an ATI Radeon 9800 Pro (chipset: ATY,R350). My
> Nvidia card is a GeForce FX Go5200. Both are running OS X 10.4.7
> (PowerPC).
> 
> Also, I did try disabling GL_DEPTH_TEST for the backdrop text just to
> see what would happen. Both rendering problems did go away so I'm
> pretty confident this is indeed a z-fighting problem. The unfortunate
> side effect is the text doesn't get correctly culled or drawn when
> below/ontop of the Earth model.
> 
> So I'm thinking maybe in addition to a tweak API call, we might
> want/need a render-method API tweak. Down the road we could introduce
> this current render system (as default), a mode that disables
> depth_test internally (for text that doesn't need to care about the
> side effects), and a stencil buffer mode.
> 
> Or I don't know if there is a bug report channel to get ATI to fix
> this problem. We need to craft a good bug report to submit though.
> 
> Thanks,
> Eric
_______________________________________________
osg-users mailing list
[email protected]
http://openscenegraph.net/mailman/listinfo/osg-users
http://www.openscenegraph.org/

Reply via email to