Personally, I think nVidia's got rocks in its head to come off of their original recommendation of a 480W PS for the 6800U, as that's a high-priced, upper-end 3d card that will be sold in limited quantities compared to other products in the nV40 family.
First, there won't be any OEM anywhere, I would bet, who will ship them in systems with only one molex conntector attached. (I mean, what would they do, include a note with the system that says, "If you try and overclock your $500 3d card, or run intensive 3d games with it, you will need to manually attach the second molex connector, and may need to replace the current 350W PS you purchased with a 480W model for reliable operation"...? That would go over like a lead balloon.)
At home I'm running a Barton 2500+ with 1 gig of pc2700 ram (2x512) overclocked to a conservative but completely stable fsb of 180MHz+ (365.xxx MHz DDR) in dual channel mode, in an nf2-400 Chaintech 7njs motherboard, with onboard sound and onboard SATA disabled, a Promise TK2K PCI PATA RAID controller paired with two WD 100 0JB's in RAID 0 (8-meg cache), and a pair of WD 100 OBBs on the onboard IDE PATA primary channel. Secondary onboard IDE supports a DVD/CD-R, and a CD-R/CD drive. I'm running a stock-clocked 9800P in the AGP slot, and an Audigy 2 Platinum PCI soundcard. I'm fully networked with the 7njs's onboard network controller, and I use a USB keyboard and mouse, and have several system fans of various types in an aluminum Lian Li PC62 extended chassis case (which I dearly love because of its depth.) I also have a firewire card in there somewhere, but I've disabled it, IIRC.
Anyway, powering the whole thing is an Antech TruePower 430W, and although I know I have some room, I wouldn't want to push the 430 much further, especially on the 12V rail (often consumed by drives.) I don't think this is a wise move by nVidia--to come off of the original PSU requirements for the 6800U, which is a high-end, specialty 3d card to begin with, and a product that will not interest anyone else except "enthusiasts" in the first place. OEMs aren't likely to be influenced by this latest nVidia statement anyway, since they'll do their own in-house testing to determine what works with the 6800U and what doesn't in the way of their own PSU's included in the systems they sell, regardless of the generic wattage nVidia "recommends." That fact, among other things, really has me curious as to why nVidia didn't let well-enough alone with the original recommendation of 480W PSUs for the dual-molex 6800U.
More puzzling still is that I read a bit more "in depth" about this from some nVidia employee other than JHH, posted recently at Gamespot, I think, in which the nVidia employee made a very intelligent statement about the original 480W PSU recommendation. He said (paraphrased): "We recommended high to begin with because there's a difference in efficiencies offered in PSUs, between the wattage they purport to support and what they actually do support in reality, and we felt that a recommendation of 480W would keep everybody sufficiently powered." That is a sensible approach and, of course, the correct approach for the 6800U. But to come off of that now and say "~350W" is to negate all of the good sense in the original recommendation, since all 350W PSUs are not the same, either, in terms of efficiency. Obviously in certain systems configured in certain ways, with certain power supplies, 350W won't be enough for the dual-molex 6800U. So they move again from a sensible position to an insensible one, for the sake of PR bullets, and I really do think they had it right the first time.
But the most baffling presumption of all out of all of this is JHH's presumption that anyone who is not an "enthusiast" would ever be interested in the 6800*U* to begin with. I don't see that being the case, as people unenthusiastic about 3d gaming are not likely to buy a 6800U, or an x800PE, either, imo.
