New 3dlabs graphics chip/cards.

glw

Newcomer
Mentioned here (warning 9.3 MB file):

http://www.3dlabs.com/support/developer/ogl2/presentations/GLSLMasterClass2004.pdf

Code:
Future 3Dlabs Products

In the near future, 3Dlabs will be announcing products based on a new
graphics architecture

These products will be the first that we’ve designed completely since
the OpenGL Shading Language came into existence

“No Boundariesâ€￾ support for the OpenGL Shading Language

Highly parallel, scalable architecture

Continued emphasis on quality and precision

World leadership application and benchmark performance

Support for large memory configurations

More mature GLSL compiler

Driver will combine the best of Wildcat4 and Wildcat VP drivers

Much more information to come soon

I like the sound of "No Boundaries" GLSL. :)
 
Thankfully 3dlabs has been pushing full virtualisation all the way I believe in their draft specs for openGl 2.0 they require that drivers multipass any shaders that the hardware can't do ( or emulate in cpu ). I believe their last chips had texture memory virutlisaiton.
 
bloodbob said:
Thankfully 3dlabs has been pushing full virtualisation all the way I believe in their draft specs for openGl 2.0 they require that drivers multipass any shaders that the hardware can't do ( or emulate in cpu ). I believe their last chips had texture memory virutlisaiton.

Yep since the Permedia 3
 
P10 / Wildcap VP took the Virtual Memory beyond just texturing and, supposedly, anything that may reside in on-board ram could operate from system RAM (although, obviously, there are parts you wouldn't want to!).
 
I do seem to recall that Creative Labs had a plan to bring P10 technology into the consumer space back in 2002/2003. Does anyone know if this new GPU has any chance of making it into the consumer space ?
 
"Picture Perfect" went on sale in the US and was based on P9, which is a cut down P10 architecture.
 
been waiting to hear about a next generation P11/P15/P20/ P-something.

I got interested in P10 when reading that 3DLabs had 200 something SIMD processors within 76(?) million transisters. could have more or less processors depending on processing needs.

If they made a cut down P10, the P9, certainly they could make a souped up P11 in a similar fashion, since P10 is a scalable / modular architecture (just going by what ive read)...
 
The ALU's for the fragment processor were Ineger based. I'd be expecting a move to float based processing.
 
DaveBaumann said:
The ALU's for the fragment processor were Ineger based. I'd be expecting a move to float based processing.

and IIRC they were actually scalar units, batched on-demand into vector (simd) units (wich was the big breaktrhough with p10)
 
The paper very much suggests we'll see a hardware noise() function in their next architecture... nice :D
 
Back
Top