one said:Well I'm not saying it's particularly bad, just assuming it'll be OK yield if not fantastic for the comparison with RSX
I don't think it will be fantastic, but I think with ATI's approach they will be closer in cost than they would have been with a single chip design (which is only 10% bigger to begin with). I admit the GPU on the 360n will probably be more expensive, although the advantages in yields with two smaller redundant part is interesting.
ATI noted that both the Core and the Daughter Core had redundancy. So 2 smaller cores with redundancy should get better yields than 1 larger core (even if it has redundancy as well). Per wafer you should get more chips with the 2 die design (but additional costs as well). Of course any gains in die production will be offset (and then some) by packaging, shipping, etc. I just wanted to note that the GPU size differences may not be as significnant at first glance.
As for ATI, not all the next gen chips have had problems. It was noted that a couple were fine on the first spin, including Xenos. Xenos was the chip that made them confident for going 90nm in one shot... so if Xenos had issues then that makes little sense.
The other issue is Xenos is 500MHz, while R520 is predicted to be in the 700MHz range. The flip side is RSX is aiming at 550MHz, when the current G70 is at 430MHz. Now some of the extreme edition cards are all the way up to 490MHz, so it seems 550MHz should be attainable at 90nm. But higher clocks usually mean less yields as well. So it would seem the 500MHz Xenos is a pretty conservative design compared to the R520!
Last edited by a moderator: