The original source says the system requires 300W, which is what we expected, Kyle misread it to mean the card itself requires the full 300W.
BTW, where does the 300W come from when the PCIe only supplies 75W as a PCIe1.1 slot, and 75W on a 6 pin connector and then 100W on an 8 pin connector. Sounds like he's having trouble adding.
Also I find it funny that he writes about the card requiring 13+" of space (but not saying the actual card length) and implying it'll be the biggest card, yet, the GF7900GX2 would be longer if the 12.5" OEM R600 card at VR-Zone are accurate.
Looks like rumour-mongering of the InQ variety, but without the natural tendency of most people to distrust the InQ.
Must be a slow news day, especially with [H] not actually getting their hands on one, or seeing one live, unlike the EEtimes article which says 200W R600;
although being the amd/ati fanboy i am, i still have to think that, since that is either a quad-fx motherboard or a workstation motherboard the results could be slightly flogged off, just saying this is a possibility. also, i wonder what processors and power supply, since the 2x r600s will take up 600W~9% and the 2x cpus will take up 400w i'd say it would have to be a pc power and cooling 1.2kw one.looks like it has a good 4+gb of ram, lol
Hope that isn't true. That is a little to power hungry. Hard to believe from a company that is trying to save power.
yeah. maybe why they delayed it till may. I think it'll most likely have an April launch, actually. AMD has said "several weeks" and "a few weeks", and both "several" and "few" can mean "two or more". Even though you typically think of "4 or more, or at least 3" when you hear "several". Remember that PR people tend to make things like this ambiguous.
As for the card: THE 12"-13" CARD IS FOR OEM'S ONLY, NOT RETAIL, FOOLS.
The nVidia fanboys always ignore this.
Nvidia says a single 8800GTX is capable of 520 G/flops. :?
Yeah but no one's seen it in the wild, theory versus practice.
It depends on how they are implemented, from B3D's thread;
At least through CUDA and DX9/GL fragment programs, we've seen ~340GFlops for MAD on G80. We've never found the magic extra MUL, and the CUDA documentation and the CUDA forums says it isn't there, at least not for general use. R580 is ~240GFlops if just using MADs, and 360GFlops for ADD+MAD (you can do this with pain via DX/GL, but you can do it easily in CTM since you can write actual ASM). Of course, all of these shaders are contrived and your experience will vary with real code.
SO with a practical implementation of 340Gflops on the G80, to the observe 500Gflops on the demo versus the theory of 500+ in the G80 which has yet to be expressed.
I can't wait to see the final specs because the talk of 64 - 5D ALUs improved to be scalar and fully MADD, that's making it very interesting indeed.
EDIT: Something to consider, is that theory of hidden shaders, backed on that 340GFlops x 150% = ~ 520GFlops (510 really), so who knows maybe we have that to look forward too to add to all the rumour mongering of late. :twisted:
Yeah... a bit, wouldn't you think? Considering the GTX peaks at 145w / 12.2 amps. The R600 would consume 25 standalone amps - 102% increase Though I think they meant the complete configuration (all components) - heh. Yet even 200w is a hell of a bump from the GTS using 110w and GTX at 145w.