A bit of NV30 speculation from Xbit...
- Thread starter Bigus Dickus
- Start date
Think of the dorkiest name possible.
That is what marketting will eventually come up with.
That is what marketting will eventually come up with.
And the good old boys were singing 'Bye Bye Mr. ATi' 
Pixel Hoover Tarquin Overdrive with chips!!(tm)
Pixel Hoover Tarquin Overdrive with chips!!(tm)
Luminescent
Veteran
We already know most of the CineFx architecture details, have read projections of deferred rendering techniques such as tile binning to reduce overdraw, and the substantiated rumors that it will be clocked around 400MHz core and 800-1000MHz DDR II memory. An Nvidia rep even stated the NV30 is flexible enought to pack float operations and execute 2 64bit floats in the same amount of cycles as 1 128 bit float.
However, there is one thing that is driving me insane when it comes to further speculation. How many pipeplines!!! How many vertex pipes and how many pixel pipes. We know there are 16 tmu's, so if there are 4 pixel pipelines, it must be a 4*4 setup, while 8 pipes would yield an 8*2 setup (hoping for the later). Pipelines usually come in powers of 2 and usually in greater or equal amounts than the previous generation, thus NV30 will either have 2 vertex pipes or 4. To be very competitive, it should have 4, although the documents said 1.5 times geforce 4 performance (leaves me in the dark). Being left in the dark stinks. I've picked up some clues from the previous NV30 speculation threads, lets keep it coming.
However, there is one thing that is driving me insane when it comes to further speculation. How many pipeplines!!! How many vertex pipes and how many pixel pipes. We know there are 16 tmu's, so if there are 4 pixel pipelines, it must be a 4*4 setup, while 8 pipes would yield an 8*2 setup (hoping for the later). Pipelines usually come in powers of 2 and usually in greater or equal amounts than the previous generation, thus NV30 will either have 2 vertex pipes or 4. To be very competitive, it should have 4, although the documents said 1.5 times geforce 4 performance (leaves me in the dark). Being left in the dark stinks. I've picked up some clues from the previous NV30 speculation threads, lets keep it coming.
Luminescent
Veteran
Sorry Kristof, I'll edit that to "projections".
Here is an OpenGL pdf from NV News
http://www.nvnews.net/files/OpenGLforNV30.pdf
I am just a spectator like everyone else. õ¿õ
http://www.nvnews.net/files/OpenGLforNV30.pdf
I am just a spectator like everyone else. õ¿õ
Simon Templar
Newcomer
TempestX,
Not to be too critical, but they have a trademark on "Eclipse", not a patent, That is of course unsless you know something I don't!
-Simon T
Not to be too critical, but they have a trademark on "Eclipse", not a patent, That is of course unsless you know something I don't!
-Simon T
Luminescent
Veteran
Hmm, I had never read the "over" 1.5 times faster vertex performance than nv25, as stated in the document. Is this the same document that has been available on the Nvidia site for some time?
Simon Templar
Newcomer
TempestX,
LOL. There has been so much rumor that even if someone was to spill the proverbial beans I would have a hard time telling the difference between real and fake. Futhermore I can't think of an architecture that hasn't been rumored to be the NV30 a IMR/TBR, 4x4/8x2/6x2, 128/256bit, DDR/DDR2, Single Chip/Multi Chip etc. I guess only time will tell...
P.S. On a side note I checked out TESS (Trademark Search) and it doesn't seem that Nvidia has any new trademarks registered yet.
LOL. There has been so much rumor that even if someone was to spill the proverbial beans I would have a hard time telling the difference between real and fake. Futhermore I can't think of an architecture that hasn't been rumored to be the NV30 a IMR/TBR, 4x4/8x2/6x2, 128/256bit, DDR/DDR2, Single Chip/Multi Chip etc. I guess only time will tell...
P.S. On a side note I checked out TESS (Trademark Search) and it doesn't seem that Nvidia has any new trademarks registered yet.
