The GT300/Fermi Thread - Part 2!

[***Deimos***]

charlie has been saying nvidia yields are bad since g92 although yields probly arent good if die size rumors are true.

http://www.nvidia.com/object/product...gt_240_us.html
http://www.nvidia.com/object/product...9800gt_us.html
340 doesnt look to be a rebrand because gt 240 is 4.8GPixels and wider bus.

die size rumours???
what is there to question??

nVidia openly disclosed that Fermi is 3B transistors.
We also know G200 was 1.4 billion transistors covering a 576mm2 die surface area built on a 65nm process.

Put 2 and 2 together. Its a virtual certainty that Fermi die size will be very large (about as big), and the corresponding heat, power, and yield issues that come with that.

Nobody in the history of silicon has doubled yields by making a 2x bigger chip.

FUN nVidia GPU CLOCK STAT:
Remember GF2.. 175, 200, 225Mhz
GF4.. 250, 275, 300Mhz
GF6.. 325, 350, 400, 425Mhz - notice a pattern yet?
GF8.. 513? 575, 600, 612, 650 - ok a few bad ones..
G200... 633, 602, 612, 648, 576, 738... umm.. what was wrong with nice even spaced 25Mhz increments?

according to wikipedia (ie 9 year old):

GeForce GTX 470 	March 2010 	GF100 	40 	3000 	576? 	1 	PCI-E x16 2.0 	1280 	448:56:40 							GDDR5 	320 	11 	3.2 		
GeForce GTX 480 	March 2010 	GF100 	40 	3000 	576? 	1 	PCI-E x16 2.0 	1536 	512:64:48 							GDDR5 	384 	11 	3.2