bdmosky said:
Now you're speculating on the overclockability of an untested core? Please be more careful here! Everyone said the 9700 wouldn't overclock well and look. Who knows. 600MHz may be possible on the NV30... we'll have to wait and see.
Um---we are talking about Apples and Oranges here...
The R300 in the 9700 Pro has a (comparatively) tiny fan, and a mediocre, run-of-the-mill heatsink. Yet, it often overclocks well even under thermal situations which are a long way from "ideal."
The nv30 at 500MHz as pictured thus far, *already* sports the largest 3D card fan and cooling solution I have ever seen--direct from the factory. The evidence is, at least circumstantially, that at 500MHz the chip is already overclocked to an average maximum (otherwise the gigantic fan would not be needed.)
It's a mistake to underestimate the inconvenience and extra expense such a fan involves. It practically kills the product dead from a system OEM perspective--as few will want something that eats PCI slots, makes a big racket, and produces a lot of heat. So clearly, putting such a fan in place on a factory model 3D card is not a light undertaking but is the result of some very serious decisions and compromises.
My own opinion is that nVidia felt an absolute priority to have the nv30 be shown to outperform the only other product in its class--the 9700 Pro. Thus, even acknowledging the negative ramifications of implementing such a HSF on a factory product, nVidia felt that the product would go nowhere if it released it running at clockspeeds low enough to require the kind of normal cooling present with nVidia's GPUs. Thus, circumstantially, it can be speculated that at a clockspeed low enough to allow nv30 to function long term with a "standard" type of cooling fan, nVidia felt nv30 either would not be very competitive with ATI's R300, or else it would not present a clear picture of performance superiority, even though it would ship 1/2 year later than the 9700 Pro.
That nVidia never, ever would have included such a fan scheme from the very beginning is evident. One of the reasons you go to a smaller manufacturing process is for lower current and cooler running properties, and also for the ability to include more circuitry in the same size die you might use with a larger process. Of course, adding more circuitry is counterproductive to power and thermal issues, and I think that's where the nv30 wound up. It needed to run faster than it was running to be competitive in the way that nVidia wanted, and so the fan concept to achieve that goal was introduced into the project. IMO. This certainly explains part of the delay in nv30 to date.
