|
Greetings All, I'll pipe in here a bit. Maybe we are looking at this from too low a level, and need to back up to a higher altitude. Take the difference between the GeForce and FX lines at Nvidia. One is for textures/polygons and the other is for pixels/triangles as the two product lines are vastly different in their target market. What will drive the projects design goals are not so much the people who will be using the card, but the applications that will be using the product. To that end, for those of you that are planning on using this card for other than its primary purpose (that being a rock-solid graphics adapter that "just works") what are the applications that you are going to be running, and what fill rate/pixels per clock/vertices per second/memory bandwidth do you need for this product to meet your needs both now and for 2 years down the road. When we have a good idea of what applications are going to be running, and what they'll need to run appropriately, then we can guage how well this card will meet those needs in its performance. Or, am I chasing a rabbit? RandyW That being said Timothy Miller wrote: So, it's been pointed out that 200Mp/s would be adequate for high framerate at 1600x1200. That seems right. So, why are all the other vendors so bent on making their GPUs run at 500MHz? I can see how you'd want the geometry engine to run at high speed, but we don't have one. Plus there are programmable fragment and vertex shaders, but we don't have those either.Exactly HOW FAST do we need to make this chip go in order for it to be viable? If we're talking about memory bandwidth, well, we've got it. The 1600Mp/s figure isn't going away. Furthermore, even at 200Mp/s, we'll still saturate that if we turn on enough features, which is the norm in games, for sure, and even still a significant issue with 2D operations. If we decide to make it run at only 100MHz, there are some adjustments I'd make to the design. For instance, when doing linear filtering, you can request the two adjacent pixels in one shot, so that's what I'd do. The result is that the rendering engine would be able to fun full-out at 200Mp/s, until you turned on so many features that even other vendor's cards would hit their limits anyway! Comments? Opinions? _______________________________________________ Open-graphics mailing list [email protected] http://lists.duskglow.com/mailman/listinfo/open-graphics List service provided by Duskglow Consulting, LLC (www.duskglow.com) |
_______________________________________________ Open-graphics mailing list [email protected] http://lists.duskglow.com/mailman/listinfo/open-graphics List service provided by Duskglow Consulting, LLC (www.duskglow.com)
