Doom3 a TWIMTBP game?

What ever became of Doom 3 and Ultrashadow support? If my memory serves, Carmack was pretty enthusiastic about the concept.

In a game as shadowy as Doom looks to be, perhaps a more efficient shadow rendering method can make up for weaker shader support.

Does ATI have something like Ultrashadow?

 

IIRC ultrashadow is nv_depth_clamp or some other implementation to cut off shadow volume generation at the far plane. I seem to recall nVidia (or JC) saying while ATi is technically capable of doing it, they need to find a different implementation (i.e. patent).

I could be wrong though, it was some time ago that I actually read that.

 

UltraShadow is GL_EXT_depth_bounds_test. ATI may be technically capable of doing the same, but only with a new chip

 

so wait--it's an extension without the nv_ bit--does that mean that any manufacturer could support it?

 

Other manufacturers can use NV_ extensions if they want to, look at 3DLabs, their cards supports low level extensions like NV_register_combiners, NV_vertex_program.

 

Are there multiple versions of the voucher or am I -err- reading-challenged? (can't find it on mine)

cu

incurable

 

I think there are different versions, mine looks different to the ones I saw when they first came out. I'd scan it but I put it down somewhere in the last 24 hours and have nfi where that might be. I remember seeing pics of white vouchers, while mine is grey and black, and double sided.

 

Mine is monochrome with a picture of Gordon on the front with a bunch of "Get in the Game" hype and black & white on the back with a silver scratch-off thingy. (That I still haven't scratched off yet)

Came with an AIW 9600 pro, if that helps any.

 

thanks for the refresh. Do you know what ATi is currently lacking in implementation?

 

If other manufacturers want to license those extensions, yes?

 

They lifted the licensing restrictions on their fragment program extensions a little while ago after they got some heat for it, IIRC.

 

Schematically, UltraShadow works this way: You provide two reference values(min, max) which form a depth range. For every pixel/sample, the depth value in the depth buffer is checked against this range, and if it's outside, the pixel/sample is discarded.

In practice, this is mostly useful for stencil shadows. If you did that per pixel, you would only save bandwidth (best case: depth read only, as opposed to depth/stencil read and stencil write*). But even with 16 samples per clock and combined depth/stencil, you're usually not going to be bandwidth limited on a 256bit DDR bus (16 * [32 bit read + 32 bit write] = 1024 bit, compression should put that down to a third at least)

To save fillrate, you have to employ an early Z method that can discard several pixels per clock, like hierarchical Z. And you need a depth range per tile so you can check on both sides of the reference range.

AFAIK the R3x0 only stores one value per tile in the hierarchical Z buffer, and it doesn't support checking of the depth buffer content against a reference range. So it basically lacks everything UltraShadow needs.



* that's for separate depth and stencil buffers, I'm not sure the current chips handle them separately.

 

Steam got taken out earlier this week when they released an upgrade to the client. Of course, the next day they got DDOSed so that probably has an effect too...

I'm pretty sure that once you've entered your ATI Voucher into a new Steam account, ATI's obligation to you is already over, and now it's Valve's problem to get the final product into your hands.

 

That sounds logical. Now let's hope that they register that you've downloaded the game when you've finished, and not when you start, or else we might see a lot of trouble.

 

Forgive my ignorance, but how is this different from a clip plane? Place one clip plane at depth min and another at depth max...

 

This is useful for stencil shadowing, since you don't have to do stencil ops in areas that cannot be affected by a light source (occluders, attenuation).

 

Ok, I see how it's useful for occuders, but not for attenuation.

 

A surface that is far away from an attenuated light source will never be affected by it. From the position of the light source, you can derive a depth range outside of which you don' have to check whether a pixel is in shadow. It will simply always be in shadow.

 

Thank you Xmas,

You made me think of something, r3xx can only store a value instead of a range, but can't they implement just one of the depths? Say... zmax? Then just check if a pixel is beyond that and discard it or if within compute the light value?

Also, about your reply to OpenGL guy, DOOM 3 lights always have a radius, you mentioned ultrashadow works on a per-light basis, then I take it that ultrashadow uses this level information, or is it derived (as you put it) by an algorithm, if you know.

 

It looks like that once HL2 is made available over Steam, it seems like if you're marked as having "purchased" HL2, you'll have permanent access and redownloads to that title, and that you won't have to repurchase it again if you choose to redownload it (case in point: people with the HL1 pack active are allowed to redownload HL/CS/TFC/etc as many times as they want, as long as they're marked as being owners of said pack). So, even if you download it and it fails, you can just start again whenever you want because you're marked as 'owning' it.

Of course, this is just my opinion after playing around with steam for a little bit now. When they release CS:CZ it'll be the first test of Steam is a distribution model of games, as it'll have both Steam purchases and the typical box system.

If Valve gets this whole system figured out, Steam could be quite a good system. It certainly has lofty goals.