NV30 to "unlock" advanced graphic features in UT2k3?

Page 2 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

MrBond

Diamond Member
Feb 5, 2000
9,911
0
76
I heard the nv30 will not only make your games look amazing, but it will also bake you PIE. Any pie you desire, and the NV30 will bake it for you :D

Holy crap, 180mb textures is nuts. I just bought a 128mb gf4 thinking that'd be more then enough vram :(
 

codehack2

Golden Member
Oct 11, 1999
1,325
0
76
Originally posted by: jliechty
Originally posted by: BD2003
Let me get this straight, radeon has 8 pipes 1 tex. NV30 has 8 pipes 2tex. The game obviously uses multitexturing, so wouldnt the nv30 theoretically need only half of its pipelines to match the 9700's speed? Therefore, AA wouldnt be free, but itd certainly be a hell of a lot faster...
NV30 will not only have two texture units per pipeline (whether or not that will help I'm not competent enough to say), but it will also be on a lower micron process than R300, and thus (probably) be able to hit higher clockspeeds. I'm sure the higher clockspeeds will help it.

You guys are getting confused on the difference between mpixel fillrate and mtexel fillrate and the role played in AntiAlaising. Think of it like this... Pixels usually refers to values written to the frame buffer (that are the pixels shown on your screen, pixel = picture element). Texels refer to the pixels that construct a texture . The word texels comes from texture elements. AntiAlaising is a routine that is performed in the pixel pipeline before the frame is written to the frame buffer. TMU's have nothing to do with AntiAlaising. Therefore, the number of TMU's that NV30 has is irrelivent to AA performance.

jliechty, you are correct about the clock speed of the nv30. If it is indeed higher than the r9700, then that could lead to higher AA performance... via a higher mpixel fillrate.

CH2
 

Bovinicus

Diamond Member
Aug 8, 2001
3,145
0
0
Hrm, you seem to be correct that it only has one texture unit per rendering pipeline... However, in Anadtech's article it says that it can do 16 textures per pass. How is this possible with only one texture unit per pipeline?
 

codehack2

Golden Member
Oct 11, 1999
1,325
0
76
Originally posted by: Bovinicus
Hrm, you seem to be correct that it only has one texture unit per rendering pipeline... However, in Anadtech's article it says that it can do 16 textures per pass. How is this possible with only one texture unit per pipeline?

Rather than wrap my own explaination around this, I'm going to qoute Beyond 3d's R9700 Review:

"This is achievable via a process commonly termed as 'loopback' - the first texture is sampled in one cycle and the result is stored on chip, and then the second texture is read in the next clock; this process is repeated up to 16 times. The textures can be any combination of one, two or three dimensional textures with bilinear, trilinear or anisotropic filtering applied arbitrarily"

CH2

 

Gstanfor

Banned
Oct 19, 1999
3,307
0
0
Umm.. errr. Lets start at the begining...

1)" What if nVidia implemented a (transparent to applications) texture compressor (based on S3's pro level texture compression or 3dfx's compressed formats)? All of a sudden NV30 is able to fit more textures into 128mb than other cards (which would be a reason why NV30 could do this and 9700 could not)."

-You can't compress textures twice... Texture compression can already be turned on and used in UT2k3... how would NV30's so called transparent compressor be any more effecient that the current implementation, which is based off of s3's spec mind you
You need to read up on S3 texture compression. There is a professional level of compression available (but it's not one of the common S3TC formats - you have to pay to use it).
There is no reason why NV30 could decompress then recompress textures as they load (probably storing a tag with them identifying the original texture format).

Greg
 

codehack2

Golden Member
Oct 11, 1999
1,325
0
76
Originally posted by: Gstanfor
Umm.. errr. Lets start at the begining...

1)" What if nVidia implemented a (transparent to applications) texture compressor (based on S3's pro level texture compression or 3dfx's compressed formats)? All of a sudden NV30 is able to fit more textures into 128mb than other cards (which would be a reason why NV30 could do this and 9700 could not)."

-You can't compress textures twice... Texture compression can already be turned on and used in UT2k3... how would NV30's so called transparent compressor be any more effecient that the current implementation, which is based off of s3's spec mind you
You need to read up on S3 texture compression. There is a professional level of compression available (but it's not one of the common S3TC formats - you have to pay to use it).
There is no reason why NV30 could decompress then recompress textures as they load (probably storing a tag with them identifying the original texture format).

Greg

Do you have a link? I've hit via's site, along with google and have came up empty.

CH2