Originally posted by Krux
Its suposed to be a quite sulution actually so I don't think noise is a factor, they didn't want to just put a huge fan on it so they used abits heat pip tech that they designed for there geforce 4 series.
not quite m8
the gf FX is being released overclocked... the fx FLOW or whatever the cooler is called is actually quite loud per people who worked their booths during comdex... also the card sample @ the time was supposed to have been clocked far slower than the working models that are supposed to start @ 500/500 ddr II
the lack of integration of the low K process will affect the card in terms of heat as there is more resistance to the current passing through the circuitry... if they did not have a massive cooling system such as the fx FLOW they may no have been able to reach air-cooled speeds with a regular hsf of 500/500... more likely 400/400-ish would have been the upper level STABLE...
therefore there is some doubt as to howmuch oc'ing headroom this baby is going to have...
the OTES cooling system used on the gf4 series is NOT the same as the gf FX cooler even though they do look similar...
beyond this let me reiterate that the gf FX should easily be around 15-30% faster in regular resolutions... it is when things like FSAA and AA and high resolution load is on the gpu that we will find out how many frames it can push out... it has a smaller memory bandwidth and appears to push fewer triangles... the efficiency of the design should be able to overcome some of these handicaps but the tests will be the ones that show this...
the card IS massive... it DOES have a huge cooler which may require you to keep 2 pci slots since it does sort of leak over into the second adjacent pci slot as well (please look to another thread where I posted not only the video card but the motherboard as well... this is one MASSIVE card)
the card should be priced around 399 USD... the same as the retail price of the 4600 and the 9700pro... I am a bit surprised @ this since the card employs 12layer pcb (this is done to cancel out the signal noise generated by the card design and the high clock frequencies) for reference the 9700pro employs 8 layer pcb board design... the more layers... the more expensive the process... during the design stage both the cards started out with a 10 layer design... the r300 boards dropped to 8 with revisions and refinements but apparently the higher clock speeds of the nv30 do not allow this to happen...
now there are a lot of things we can speculate about the card/cards BUT without a basic understanding of how the architecture works or ddrII (which has higher latencies than ddr) and what not... a lot more rumours and inuendo will be spread...
scriptasylum... I am a fan of the tech that ati use but its less so their tech than the bs that nvidia routinely spews out that really irks me... if you search through a few threads you will see a post by me that explains my views on the different video cards/cpu's et al
so I will not go into that again
btw the next iteration in ati's lineup is the r350 which will continue to use the 0.15 process and probably be clocked @ around 400/400 ddr.. which with its bigger 256bit memory system will emmulate the gf FX's fill rate and therefore should perform @ a similar level.. and better with more eye candy turned on with FSAA and AA enabled to high levels...
I will also be looking for improved 3d IQ on the gf FX since that has been the buzz word...
now remember folks... this card that will be released is not only over 6 months late... it is also highly overclocked (read earlier statements and compare with clock speeds of the quadro FX)... it obviously was tweaked with to compete better with the 9700pro which obviously threw a wrench in the nvidia product cycle...