Call me Lamer!!!

[Tatchan]

I've just attended a press event today, but as you all know, I can tell nothing except....It's really there! It's super!

 

[Qroach]

you mean matrox?

 

[Tatchan]

I meant nothing

But I would suggest people to wait a while if they want to buy a high-end graphic card, if they're not in a hurry.

 

[jb]

Ok so how long do we have to wait can you at least hint on that? And no soon is not good enough

 

[Grall]

If it's REALLY there, that makes me think BitBoys! LOOOL

Okay, so flame me to death now if you want, I don't care.


*G*

 

[Dave Baumann]

Possibly related (but probably not) you may want to keep your eyes peeled on Friday...

 

[psurge]

MURC rumor has it that "parhelia" will be announced may 15th or 16th (can't remember exactly).

 

[Nappe1]

even I don't think this means Bitboys, but someone else that has caused some sleepless nights to me.

 

[Joe DeFuria]

Possibly related (but probably not) you may want to keep your eyes peeled on Friday...

I hope this means that ATI is ready to announce a new product tomorrow...the "what the hell is it" R(V)250 chip.

 

[mboeller]

Do You speak about the new GPU which, according to a rumour on an other board should be unveiled at the 14. May?
The rumour also says that it is indeed the new chip from matrox and it can do 8xMSAA at the same speed as the GF4Ti 4600 can do 2xMSAA !!

happy waiting


Manfred

 

[Joe DeFuria]

OK, so eaxctly how fast can GeForce4 Ti 4600 render:

Quake3, 1600x1200x32, 2X MSAA? (Looking for a fill-rate / bandwidth limited situation here.)

I've been browsing some of the more popular sites to get such benchmark numbers...to no avail. Anyone have a link?

Of course, assuming Matrox's AA technique is MSAA, we'll need to apply anisotropic filtering to sharpen those textures without increased aliasing as well. If the Parhelia performance rumor is true, looks like we might be getting a decent step closer to what I consider the "point of diminshing returns" of PC diplays:

1600x1200x32
16X "Full Scene AA" (16X on the edges, plus high level anisotropic filtering on textures, or some 16x full scene supersample AA equivalent.)
60 FPS

Traditionally, hardware developers have been trying to increase graphics power to combat two things:
1) user's desire to run with higher resulotions / AA / filtering options, and
2) Game developers desire to increase detail and complexity into gaming scenes.

Once we reach that "point of diminshing returns" however, then it's "only" a matter of hardware vendors increasing the power / bandwidth of their solutions to satisfy number 2.

Back to Matrox, it will be interesting to see what kind of product line-up they deliever. (What are the different price / performance levels.)

 

[OpenGL guy]

Of course, assuming Matrox's AA technique is MSAA, we'll need to apply anisotropic filtering to sharpen those textures without increased aliasing as well.

I am confused by this statement. MSAA has no impact on texture quality whatsoever. If the application's textures looked blurry before, they will still look blurry. If they looked aliased before, they will still look aliased. Really, it's up to the application to make sure they use good textures and good mipmaps and good use of texture filtering.

 

[jb]

Maybe,

say if they use Alpha Textures then to get them to work "right" in a MSAA you should use a Alpha blend. This however cases them to blur slighlty...I know its dose not happen that often....

 

[OpenGL guy]

say if they use Alpha Textures then to get them to work "right" in a MSAA you should use a Alpha blend.

If by alpha textures you mean textures where alpha is either 1 or 0 and alpha test is used (I'd call them cut-out textures, but that's me), then there won't be any difference for MSAA. Since each pixel only gets one texture value, you either pass alpha test or you don't. Blending isn't going to change the result since the subsample will either be opaque or fully transparent.

 

[MikeC]

Quake3, 1600x1200x32, 2X MSAA? (Looking for a fill-rate / bandwidth limited situation here.)

I've been browsing some of the more popular sites to get such benchmark numbers...to no avail. Anyone have a link?

That's why you need to rely on fan sites

http://www.nvnews.net/previews/geforce4/page_4.shtml

However anisotropic performance has improved on the GeForce4 since that preview. Also, the image quality comparison on that page is obsolete and I need to make a note of it. For a better comparison using recent drivers, check out the GeForce4 Ti 4200 preview.

http://www.nvnews.net/previews/geforce4_ti_4200/

 

[jb]

If by alpha textures you mean textures where alpha is either 1 or 0 and alpha test is used (I'd call them cut-out textures, but that's me), then there won't be any difference for MSAA. Since each pixel only gets one texture value, you either pass alpha test or you don't. Blending isn't going to change the result since the subsample will either be opaque or fully transparent.

Yes that is what I was getting at but not really related to your topic. Sorry should have made it a side note. But for the current GF3/4 users Alpha (or cut outs) are not AA-ed at all. One quick fix was to use an Alpha Blend instead of the hard test. This will give the appearnce of AA, but will also blur the textures.

 

[Joe DeFuria]

I am confused by this statement. MSAA has no impact on texture quality whatsoever.

I know I was't terribly clear. What I meant was, that we not only want to apply MSAA to the scene, we also want to apply anisotropic filtering to increase texture sharpness compared to the "reference" bi-linear or tri-linear texturing scheme.

Based on the rumor, Parhelia can do 8X MSAA at the same performance level as GeForce4ti-4600 does 2X MSAA. But I also want to know what kind of performance the Parhelia will get when anisotropic filtering is applied in addition to MSAA, because I want to be able to use both. I know MSAA doesn't "degrade" texture quality, but it doesn't enhance it either.