This has actually been covered a number of times.
The memory runs so much faster because the memory timings are much higher than used on a PC. You might be surprised at how high the latency settings are on a video card. However, this effects a video card much less than system memory and also has to do with why the memory can run so much quicker.
1. Traces between the VPU/GPU and memory on a video card are much shorter, and the memory is directly attached to the PCB. The traces on a motherboard are much longer and the memory is attached to a PCB that is socketed. This increases signaling and ground bouncing problems.
2. Memory reads from video memory are very predictable and nowhere near as random as system memory accesses.
Video processors don't run anywhere near the speed of CPUs because of two main factors.
1. They have substantially more transistors than a CPU. An AMD Athlon XP w/512K cache only has about 1/3 the number of transistors, and a P4 w/512K cache has less than 1/2 the number of transistors as the ATi Radeon 9xxx and GFFX 5xxx series card.
2. Due to the very quick update time in these chipsets the designers go more parallel in design, rather than longer development times where more optimized designs can be handled.




Reply With Quote