Nvidia would not come to market unless it could compete. Otherwise it would be better dropping the gtx285 to 40nm and upping the clock. I'd put the performance at between a 5870 and 5970 and price between them to. Nvidia could get away with a slight mismatch in performance, say 15% faster than a 5870 (rather than the 40% I expect), because they are now targeting Tesla worstations. But that is a double edged sword, if the system is awsome in Tesla, than they would fold like a dream and so the folders would love the buggers. Eitherway the industry WILL get hyped about the Fermi, if only due to it's chequered history. It will get alot of news. Thats is bad for ATI (unless it sucks, but see prvious comment about a gtx285 40nm) SO. ATi will have to "piss on the parade", so expect a release from ATi. IMO, the worst case for ATI is a single fermi being faster than a 5870. If Nvidia release an X2 fermi, ATi will have not match, Nvidia wins the performance crown. But, until it get released, it's all speculation. I'm niether pro Nvidia or ATi. One rig is gtx280 the other is 4870 with an x2 I get what ever offers the best performance within my budget. Nothing sucks like a Fanboy who refuses to see the truth. And the truth is.... We don't know what fermi is going to do. So ranting for fermi or cypress at the mo is just foolhardy.
based on the current rumours this would be impossible as the GTX480 alone is gonna be pushing the 300w limit.
The 8800 Ultra sold like hotcakes. Without benches, we can't say exactly how fast Fermi will be - but we all know that it had better be effing fast to make up for this price, not to mention the delay.
a dual chip fermi is not power feasible, we know that is true - the die size is too big and power consumption is too high, so unless nvidia can make a single 480 faster than a 5970 then they are SOL...
I think it is fun to read all these discussions, a lot of people are contributing with interesting input. It seems like everytime something new is heard about Fermi, a new discussion pops up here. It is funny how you guys are all wound up, a month before "launch", well done nVidia marketing.
except that the majority of people are pissed at nvidia and wont buy fermi. the portion that arent annoyed were gonna buy it anyway.
Not true. If Nvidia can make tri SLI'd 480s faster than crossfired 5970s then they are safe. I'm pissed off with Nvidia for delaying these cards and for their poor decision about PhysX (disabled if an ATI GPU is present)...and yet I'll still be buying their cards.
My post from a month ago mentioned that Fermi will be intentionally hampered for folding and regardless of performance in existing games it is going to be one massive and hungry monster of a card that's expensive to manufacture. I certainly wouldn't wait a month for the launch if I wanted a card now. Prices will surely be sky high, availability very low and I'm sure nVidia will want to get a highly revised edition out esp for the mid-range (and below) and that's likely to take some time, several months IMHO.
Lucky for us, that is Fermi (Quadro's/Tesla), and not the GT200 chip (Geforce GTX 400 series). The GT200 architecture is a Fermi architecture chip with a lot of features cut out (features absolutely not used in gaming), and a lot of error correction components removed, less or none clock cycle converters, less reliable boards used (where the GPU it self sits on (passing from 99.99% reliable down to 97-98%), not as good board assembly technique used. If you panic, don't... because this is what ATI and Nvidia been doing since day 1 of their first product release. This is why Quadro's/Firepro are more expensive. This means: - Much cheaper to produce. - Faster and easier to produce. - Smaller die - Less heat - Harder overclock-ability (due to less clock converters) Also the drivers of the Geforce/Radeon are less tested than their Quadro's/Firepro counterpart, to again cut cost.
please tell me what system can run 3 x 300 watt video cards + an i7 and the opther essentials without having 2 power supplies and two seperate home power breakers powering each PSU? oh right you cant give me a system because it cant be done. Tri SLI will 100% require 2 PSUs and an extension cord running across your house so you dont blow the fuse/trip the breaker to the outlet by the PC.
GPU Cards are almost like Honda. They like the letter X and they also love to put random consonants before said letter to make the product sound cooler. Heh heh.
The fact that nVidia calls for Fermi certification of cases should tell you quite a lot about Fermi's thermal envelope. Fermi X2 can only be considered wishful thinking at this point. Hell, even Fermi is still wishful thinking (for some ).
3*300W + say, 400W for an overclocked CPU+mobo = 1300W - a bit on the high side but doable on one PSU. In regards to needing two power breakers, an average kettle uses 2.5kW and still doesn't blow fuses. Of course, we've still to see actual power usage figures, so let's hope they aren't anything like 300W.
If you think about it, for twelve years ago those graphics are pretty good. I certainly would have been pleased with them (i was only 4 so would have settled for low standards)