Something wrong with the HL2 Story

MistaPi said:
Is there some way centroid sampling can be forced in the driver?

As per all the other threads, it's possible to likely for the R3x0 cards, not possible for NV3x.

You can still turn on AA for the current cards, but it will produce artifacts. The only alternative is to crank up the resolution and live with it.
 
Bouncing Zabaglione Bros. said:
MistaPi said:
Is there some way centroid sampling can be forced in the driver?

As per all the other threads, it's possible to likely for the R3x0 cards, not possible for NV3x.

Thanx. I must have misread Valve statement. I thought they said drivers aren't likley to fix the problem period. :oops:

I know centroid sampling does not apply for NV3x.

You can still turn on AA for the current cards, but it will produce artifacts. The only alternative is to crank up the resolution and live with it.

That I got. :)
 
MistaPi said:
Bouncing Zabaglione Bros. said:
MistaPi said:
Is there some way centroid sampling can be forced in the driver?

As per all the other threads, it's possible to likely for the R3x0 cards, not possible for NV3x.

Thanx. I must have missread Valve statement. I thought they said drivers aren't likley fixed the problem period. :oops:

Drivers *alone* can't fix the problem. ATI R3x0 cards have the hardware on the chip to do centroid sampling, but it isn't currently exposed in any API. This is why it is possible for some kind of future fix for ATI cards if they can force centroid sampling for HL2.

Nvidia NV3x cards don't have the centroid sampling hardware at all, so they won't ever get centroid sampling by enabling anything in their drivers.

HTH.
 
I'm wondering if IHVs will simply offer a SSAA quick-fix in their drivers when HL2 appears; perhaps only up to 2x level or something like that. It's obviously not a complete solution, and may well just turn out to be as bad as using MSAA, but for the top end cards, using 2x SSAA may be better than nothing at all.
 
Something ATI could do to 'hack' in centroid support without requiring API modifications (but would still required dev support) is to create their own FOURCC texture formats for the various required texture formats, that when used will force centroid sampling.

An example would be something like

CA32 = ARGB 32 with Centriod Sampling
CDX1 = DXT1 with Centriod Sampling
CDX3 = DXT3 with Centriod Sampling
CDX5 = DXT5 with Centriod Sampling

Or whatever. Sure, it's a hack, but if Microsoft doesn't want to play, then a hack, maybe what's required.
 
Colourless said:
Something ATI could do to 'hack' in centroid support without requiring API modifications (but would still required dev support) is to create their own FOURCC texture formats for the various required texture formats, that when used will force centroid sampling.

An example would be something like

CA32 = ARGB 32 with Centriod Sampling
CDX1 = DXT1 with Centriod Sampling
CDX3 = DXT3 with Centriod Sampling
CDX5 = DXT5 with Centriod Sampling

Or whatever. Sure, it's a hack, but if Microsoft doesn't want to play, then a hack, maybe what's required.
He's on fire! With this suggestion, you're sure to be on track for a free 9800. (That was you with the V5-6000, wasn't it?)
 
What Colourless says seem to indicate that it is legal for IHVs to add such new formats....is that right? Why couldn't nVidia add a new format for shadow buffers instead of special casing another format, then, or do I misunderstand something?

In any case, why couldn't a combination of application detection by IHVs and color value tests by HL2 with requested AA modes to determine in game options offered (i.e., 2x, 4x, etc) be used? Like determining the list of AA options offered by a "test for AA" option that conducts an AA test that compares to color values that would result from centroid sampling, maybe that keys to driver version, and/or even use of Steam for delivering IHV AA lists to do it for the user? Detecting PS 3.0 could override offering this test being required to offer AA, and super sampling AA options could reported regardless. Also, application side supersampling seems a feasible option for valve to offer as well (mostly for LCD usage and perhaps for HL2 mods added later that are less demanding as far as shaders).

This seems fairly simple...am I missing something obvious (or not so obvious)?
 
It certainly is an "api legal" way to do it, but it is akin to doing an OpenGL extension. And we certainly don't want NVidia, ATI, and everyone else inventing oodles of texture formats just to get around a problem in the API.

MS should bite the bullet and move centroid sampling from PS3.0 to something like PS2.1 and release a DX9.1 update with it. Then ATI can ship a driver supporting it, and dev's could do it in an abstract fashion, and we wouldn't need any hacks to auto-transcode textures.
 
MistaPi said:
Is there some way centroid sampling can be forced in the driver?

Yes, the drivers could app-detect HL2 and then force centroid-sampling MSAA everywhere. The problem, as I understand it, is that this will produce artifacts of its own in places where centroid sampling should not be used. (I believe Valve has mentioned it as causing problems with text; dunno if this refers to menus or HUDs or what.)

The ideal solution would be for the application to request centroid multisampling AA only for such objects as can make use of it. The problem is that there is currently no way to do that in Direct X, short of PS 3.0.
 
Dave H said:
MistaPi said:
Is there some way centroid sampling can be forced in the driver?
Yes, the drivers could app-detect HL2 and then force centroid-sampling MSAA everywhere. The problem, as I understand it, is that this will produce artifacts of its own in places where centroid sampling should not be used. (I believe Valve has mentioned it as causing problems with text; dunno if this refers to menus or HUDs or what.)

The ideal solution would be for the application to request centroid multisampling AA only for such objects as can make use of it. The problem is that there is currently no way to do that in Direct X, short of PS 3.0.
Hopefully they're disabling multisampling for their text anyway, so the driver could force all multisample requests to use centroid sampling and the control panel be set to app preference.
 
Colourless said:
Something ATI could do to 'hack' in centroid support without requiring API modifications (but would still required dev support) is to create their own FOURCC texture formats for the various required texture formats, that when used will force centroid sampling.

An example would be something like

CA32 = ARGB 32 with Centriod Sampling
CDX1 = DXT1 with Centriod Sampling
CDX3 = DXT3 with Centriod Sampling
CDX5 = DXT5 with Centriod Sampling

Or whatever. Sure, it's a hack, but if Microsoft doesn't want to play, then a hack, maybe what's required.

FYI, Colourless, I believe that this (or something very close) is actually the solution that may have been put in place.
 
Re: hELLBINDER

palmerston said:
But its all senseless talk anyway. My original thread does not argue whether r3xx may run hl2 better. Thats been argued to death on other threads. My point is that its a poor business decision by valve to alienate the vast bulk of their target customers by optimising for a minority product.

Lets also not forget that there are *some* consumers with kyro or matrox products, -what about them too?

If valve no longer supports opengl would a port be possible?

Valve should think more than just supporting ati, feeling sorry for the under dog is one thing, letting down most of your customers is another. Assuming of course that your customers, in this case nvidia users would like to have fsaa support.

What were valve thinking of?

I know what Valve is "thinking of"--it's your thoughts I can't figure out...;)

Well, you go right ahead and feel "alienated" all you like--meanwhile, I'll enjoy the game with my R9800P...;)

You are just kidding yourself if you think ATi is in any way the "underdog" in this situation. But I'm sure Valve has indeed "felt sorry" enough for nVidia owners who refuse to upgrade their 3d accelerators (out of some sort of misguided prejudice about nVidia products in general), and so the game will allow people with products based on older, inferior technology to "dial down" the IQ settings so that the game will "run acceptably" on their less-capable hardware. I think you have nothing to fear...;)

I say--"Three Cheers for Valve!" It's great to see companies push the envelope for a change--instead of defending yesterday's status quo. I expect that HL2 will be this year's runaway best seller--and if the game is as good as its ancestor, maybe one of the best-sellers of all times. When ID software began selling Quake, 3d gamers didn't let old 3d hardware and low framerates stop them from buying the hardware they needed to fully enjoy the game--I see no reason to expect HL2 will be any different, except conceivably a much-better seller.
 
Valve should think more than just supporting ati, feeling sorry for the under dog is one thing, letting down most of your customers is another

So instead they should feel sorry for Nvidia for not producing a card that will run an industry standard feature set at acceptable speeds? Valve is not held responsible for how Nvidia makes their architectures. Valve isnt letting down customers, nvidia is
 
DaveBaumann said:
FYI, Colourless, I believe that this (or something very close) is actually the solution that may have been put in place.

Are you sure? They said they fixed FSAA on all cards.. which to me says even stuff like Geforce1 level cards, that obviously don't have any centriod sampling HW in them.
 
Back
Top