A question on an old piece of hardware

[Murakami]

Simple question: NSR in GeForce2.. does it exist? And if yes.. is it not already present in GeForce 256? I don't know a single special effect possible in GeForce2 that GeForce1 is not capable of, including DOT3 bump mapping (per pixel lighting).. what do you think? Thanks all.

 

[DemoCoder]

NSR was in GeForce1, it just wasn't publicized. When the GF2 came out, they needed a "new" feature to brag about (cause the T&L was the same), so they hauled out the register combiner's extension, gave it a fancy name, and started promoting it.

 

[Murakami]

NSR was in GeForce1, it just wasn't publicized. When the GF2 came out, they needed a "new" feature to brag about (cause the T&L was the same), so they hauled out the register combiner's extension, gave it a fancy name, and started promoting it.

I'm agree with you.. i suspected this sort of marketing trick.. thanks a lot.

 

[Tygrus_Artesiaoa]

Also a tidbit...

gf1 could do one trilinear texel per pipe instead of how gf2 did 2 bilinear texels per pipe as default
since gf1 had a 4x1 and gf2 4x2

 

[KimB]

Re: Also a tidbit...

gf1 could do one trilinear texel per pipe instead of how gf2 did 2 bilinear texels per pipe as default
since gf1 had a 4x1 and gf2 4x2

Right...but that benefit never really materialized in the real world, as the GeForce2 was far too bandwidth-limited for it to make any difference for the part.

Anyway, the NSR was essentially new functionality exposed by nVidia that also became available in the GeForce 256. (And it really wasn't *that* new...it basically allowed the software to do all the things it did before, but in a more flexible manner...which, if I recall correctly, could essentially do more things in a single pass...sort of a pre-DX8 pixel shader).

 

[Humus]

Comparing the register combiners on the GF2 with pixels shaders is a little misleading, with only two general combiners + 1 final combiner and only two simultaneous textures you don't have that much flexibility beyond standard multitexturing.

 

[KimB]

Comparing the register combiners on the GF2 with pixels shaders is a little misleading, with only two general combiners + 1 final combiner and only two simultaneous textures you don't have that much flexibility beyond standard multitexturing.

Right...but the DX8 1.0 and 1.1 pixel shaders weren't all that powerful, either (no dependent texture reads, if I remember correctly). But those first pixel shader versions were essentially a natural extension of the register combiners.

 

[Hyp-X]

NSR was in GeForce1, it just wasn't publicized. When the GF2 came out, they needed a "new" feature to brag about (cause the T&L was the same), so they hauled out the register combiner's extension, gave it a fancy name, and started promoting it.

They did the same thing with Shadow Buffers when releasing GF3Ti cards.
(Works fine in GF3 classic.)

They also 'introduced' the 4xS antialiasing type with GF4 cards. (The GF3 cards can do it, but it's not enabled in the control panel.)

Suprisingly when they remove a feature (like RT-patches from GF4Ti), they do not advertise it that much...

 

[Hyp-X]

Right...but the DX8 1.0 and 1.1 pixel shaders weren't all that powerful, either (no dependent texture reads, if I remember correctly). But those first pixel shader versions were essentially a natural extension of the register combiners.

PS 1.0 can support dependent texture reads (texbem, texreg2ar, texreg2gb and the matrixing ones).

And the PS is not an extension of the register combiners. While it contains more (the texturing instructions), it's color/alpha instructions provide only a subset of the register combiner functionality.