megadrive0088
Regular
as you all know, in the recent Nvidia conference, the well spoken CEO
mentioned they're working on something twice as fast as GFFX for, for the laster part of this year. of course its NV35. assuming that it is indeed NV35..probably to be named GeForceFX2.. what does twice as fast imply?
to me, it implys one or all of the following things:
*twice the vertex shading performance with more vertex math units.
ala NV20>NV25. since NV30 is highly scalable and modular according to various articles, this should be relatively easy. even easier than adding a second vertex shader to NV20 to make NV25. twice the vertex math units
as GFFX would be a nice help.
*256-Bit bus- twice the bandwidth. pretty straight forward here. it would be almost laughable if NV35 stuck with 128-bits
* 2nd TMU per pipe.
yeah this has been speculated on endlessly for R300, NV30, R350 and now NV35. it seems very unlikely. but this would give Nvidia twice the texturing power in older games. but it probably wont have it. it seems
the 'more TMUs is better' war was finished with Parhelia.
things that CEO probably did NOT mean:
*twice the pipelines. no way NV35 is going to be 16 pipes
*twice the MHz for the core. no way is NV35 going to be 1 GHz
I believe NV35 will be an 8:1, running at 550-600 Mhz with twice the vertex math units, a 256-bit bus, DDRII, and probably the same shader
version as NV30. thus no V.S./P.S. 3.0. - twice the bandwidth and twice the vertex power would be enough to say that NV35 is twice as fast as NV30 as far as Nv is concerned.
mentioned they're working on something twice as fast as GFFX for, for the laster part of this year. of course its NV35. assuming that it is indeed NV35..probably to be named GeForceFX2.. what does twice as fast imply?
to me, it implys one or all of the following things:
*twice the vertex shading performance with more vertex math units.
ala NV20>NV25. since NV30 is highly scalable and modular according to various articles, this should be relatively easy. even easier than adding a second vertex shader to NV20 to make NV25. twice the vertex math units
as GFFX would be a nice help.
*256-Bit bus- twice the bandwidth. pretty straight forward here. it would be almost laughable if NV35 stuck with 128-bits
* 2nd TMU per pipe.
yeah this has been speculated on endlessly for R300, NV30, R350 and now NV35. it seems very unlikely. but this would give Nvidia twice the texturing power in older games. but it probably wont have it. it seems
the 'more TMUs is better' war was finished with Parhelia.
things that CEO probably did NOT mean:
*twice the pipelines. no way NV35 is going to be 16 pipes
*twice the MHz for the core. no way is NV35 going to be 1 GHz
I believe NV35 will be an 8:1, running at 550-600 Mhz with twice the vertex math units, a 256-bit bus, DDRII, and probably the same shader
version as NV30. thus no V.S./P.S. 3.0. - twice the bandwidth and twice the vertex power would be enough to say that NV35 is twice as fast as NV30 as far as Nv is concerned.
I wonder if we will see this new card at the end of this year, or more likely in the middle of next year.
