Not sure why Rroff has come up with the idea GT212 was supposed to be up against Evergreen, it wasn't. It was supposed to be out a LONG time ago now, really over a year now, it was a very minimal die shrink, it wouldn't have been about clock speeds but about making it profitable at the same price. It almost certainly would have been a simply replacing of the GT200 with a more profitable version of the core that was maybe 60% of the size of the GT200b and would have been the midrange card with Fermi above it.
The GT212 wasn't a massively faster core, nor was it designed as anything but a 240 shader design, it was a like for like replacement and theres no way in hell, not a single tiny chance that Nvidia planned such a massive speed bump on GT212 that a card with the same number of shaders would be considered "next gen".
Fermi was always meant to go up against Evergreen and they had probably deluded themselves a couple years ago when their bought Physx that making a gpgpu type "gpu" would be the war forward as Physx would be soooo massive , back when they started designing this thing would have been 2-3 years ago.
If Rroff seriously thinks Nvidia meant a 240 shader GT212 was supposed to be the next gen card and they pluked Fermi out of a different idea and shoehorned it in........ well.
Its just a bad design in terms of knowing what they need, what to waste transistors on and what can be made on the process.
As for minimums, you can certainly take note of them and do some calculations in your head, the 5870 isn't massively higher max fps, so for it to still have a higher average fps with a minimum of 4fps, its going to be an incredibly small percentage of frames going that slow or the average would have been dragged down much further.
I did actually think it was doing to a specific memory limit, but frankly with the last 5-6 years of AMD having significantly better memory compression and optimisation than Nvidia that doesn't actually seem all that likely if infact these are comparisons with a 470gtx with only 1.2gb memory.
Though it might be possible Nvidia managed some uber optimisations to cut down memory usage in specific benchmarks just to create a few artificial situations but again, seems unlikely.
It still won't matter end of the day, I just can't see how Nvidia can financially afford to produce 100k's of these if each takes a hefty loss. Which really points to a limited run, massive prices and zero availability after a very small amount or sold out in days, which really means as I said, it could be 100 times faster than the 5870, the fact you can't buy it and they won't make it, means its speed and price is really rather irrelevant to all but a few thousand that will get them.
lol
