overclocked said:
I´m wondering if you think we will see "big clockspeeds" in mainly mem?
800MhzGddr3 and maybe 550 for the core..
Also is it using low-k and NOT being manufactured at IBM?
Oh and btw has it already taped-out?
This seems to be an ongoing thing with respect to nVidia gpus released in the last 18 months...
People imagine that just because the current pre-market nVidia gpu doesn't top the current pre-market competition that nVidia can simply scrap it, wave a magic wand, and go to "nV45" which will solve all of their problems...Man, talk about delusional thinking....this is something I've never understood.
First of all, nV40 has yet to ship, and people don't seem to understand that nV40 is nVidia's best effort to date--it's quite simply the best they can do, even in a prototype sent to a handful of sites for reviews. nV40 as currently reviewed represents "all" of the marbles nVidia currently holds.
Assuming that nV40U is something nVidia will have no problem producing in quantity, which of course remains to be seen, it's nV40 nVidia is concentrating on right now--not nV45. Apart from possible yield issues with nV40, what's wrong with it? Is the fact that it doesn't run rings around a competitor's offering to be considered something "wrong" with it? I would say "no." Right now, the challenge ahead for nVidia with nV40 seems to be the same one it faced with nV30--yields. Let's relax for a few months and see how they do.
nV30 was basically an nV2x kludge with lots of stuff shoehorned into it. It was so bad in terms of yields nVidia *had* to cancel it prior to mass production--they simply could not mass-produce it reliably at its advertised speed of 500MHz. nV35 was the result of the nV30 failure, but nV35 itself was plagued by yield issues for most of the year last year, which in turn brought nVidia to nV38, not to mention clock reductions relative to nV30 in the interest of yields.
nV40 is a lot cleaner design, and much less of a kludge, than was nV30, it appears to me, so I very much doubt nVidia will rush into "nV45" anytime soon unless nV40, like nV30, experiences a failure because of yields. I don't see this happening because nVidia has already adjusted the nV40 production clock downwards for the sake of yields, and I think they will do it again if need be. So, let's wait and see if they can produce nV40 with acceptable yields before jumping the gun in making assumptions about nV45. I would bet that's exactly what nVidia is doing at the moment--concentrating all of its resources on nV40 production.
