So the GTX with 6 quads with 2 ALUs has fewer transistors than an X1800 (counting methods aside). Regardless, an equivalent GTX in terms of die size on 90 nm would ostensibly be smaller than an X1800 and its 4 quads. And although there have not been apples-to=apples comparisons, in situations where fillrate is similar (XL vs. GT), the 7800 holds its own.
So what are the extra transistors or die space going towards on the X1000 series. Could it be cache? The improved dynamic branching (seems like an expensive price to pay)? Does not including partial percision increase transistiors (I doubt it but what do I know)? Or has NVIDIA just done a better job with layout and circuit design ( wouldn't be all the surprising since this is probably one reason they had relative success on .11 on such products like the 6600 GT and the Go 6800 Ultra)?. Maybe some combination of these or some other factors?
EDIT: This thread may be similar to the the R520 die size and transistor count redux, but the intent is to foster discussion of the differences in architectures, and in particular, why NVIDIA's design is seemingly so much more efficient.
So what are the extra transistors or die space going towards on the X1000 series. Could it be cache? The improved dynamic branching (seems like an expensive price to pay)? Does not including partial percision increase transistiors (I doubt it but what do I know)? Or has NVIDIA just done a better job with layout and circuit design ( wouldn't be all the surprising since this is probably one reason they had relative success on .11 on such products like the 6600 GT and the Go 6800 Ultra)?. Maybe some combination of these or some other factors?
EDIT: This thread may be similar to the the R520 die size and transistor count redux, but the intent is to foster discussion of the differences in architectures, and in particular, why NVIDIA's design is seemingly so much more efficient.
Last edited by a moderator:
