> > Maybe i am on error with you, and you might want to enlighten > > me what successful computer projects on earth are a result of > > your bold mind. (dont exclude computer projects for space crafts.) > > Now the above is something a bit 'trollish'. Kindly address > the arguments, not the arguer.
Sorry, missing your answer on your expiriences, still. (i was just a bit keen on you) > A GeForce 3 produces about 88 million triangles/second maximum. Am i allowed to assume some 4 bytes per x/y/z coordinate each, some 1 bytes per r/g/b color, some 4 bytes per each of the three normalized face vector coordinate? Then i sum up to: 4*3 = 12 1*3 = 3 4*3 = 12 total: 27 bytes per triangle geometry (with optimum strips) The triangles are no lesser than 5 pixels per Triangle in benchmarks... This means at least 5*4bytes = 20 bytes of data. so i do see a 27 bytes load and 20 bytes store task for your CPU making a total of 47 bytes via the bus, assumed if triangle could get rendered directly without any further memory access. You need at least 88*47 = 4,1 GB/sec transfer rate to only first get and then get rid of the data. Now you know what you would burden the CPU bus in just a "fill up the memory and dont ask much" situation But now think that: you have 8 light sources (specular, highlight, abient nicely mixed), some 3 to 8 clipper planes, an exponentional fog function applied, you are using two sided triangles and of course misc material sepcifications. This will be done nicely by the grafics core pipeline, but you will never get it done with your estimated 34 clocks on your imaginary CPU and not with even if you double that value by the CPUs clock. > Now take a > 3GHz Pentium 4, just a few months down the road. It has > available 3 billion > clocks each second. This allows about 34 clocks per triangle. With a > properly optimized algorithm, using SSE and SSE2 to work on a > tile at a > time, this can be achieved. What the main processor lacks in physical > resources it can make up with a higher clockspeed. I believe > the GeForce > 4Ti is at 300MHz now, is it not? Fillrate is still an issue, but will > become less of one as time goes by. Memory bandwidth is ever > increasing. > Use a scene-capture algorithm and you reduce the required fillrate > dramatically. A grafics processor always has two ports by design, so it has always an advantage over a CPU design. As seen above the bandwidth is a critical factor, but its much more the complexity of a scenery that doesent ever make it desireable to put such render tasks onto a CPU. As i told you further, a balanced system utilizes any part equally. So that means the AGP bus is at nearly 100% usage for supply of data whilst the interface to the framebuffer is spent up at the same time. Its not important how fast the clock of the grafics chip is, but only that this chip can process the supplied data fast enough for the supplying bus speed. And this should be the case for nearly any sort of switchable features. Regards, Alex. _______________________________________________ Dri-devel mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/dri-devel
