[Nvida] [Tweaktown] GK110 maybe = 7 Billion Transistors

[Jaydip]

A very important question is if we recompile the existing cuda applications in Kepler do we get the added performance or do we need to use the new instructions specifically for Kepler?

 

[Keysplayr]

I'm looking forward to seeing what Nvidia and Stanford come up with for a Kepler F@H client.

 

[Jaydip]

Yeah me too, not because i run F@H but if they can come up with something innovative it will great.

 

[Alaa]

GTX 680 Transistor Count 3.5B x 2 = 7B

 

[Keysplayr]

GTX 680 Transistor Count 3.5B x 2 = 7B

Absolute possibility. I think a 6Billion Xstor Big Kepler is also a possibility.
Ah, who knows.

 

[Ajay]

I'm looking forward to seeing what Nvidia and Stanford come up with for a Kepler F@H client.

Yeah, IIRC, F@H only uses SP - so it should fold very fast!

 

[tviceman]

Starting to think that the highest end first generation big Kepler will have fused off parts, just like with gtx480. Not entirely because of manufacturing problems, but simply to maximize yields and sales - Nvidia has stated that the demand in the professional space is expected to outstrip all of their previous generation tesla lines COMBINED. For that reason, I think Nvidia will be getting big Kepler out sooner than they internally anticipated, and will respin/retweak big K to release fully unlocked chips for second gen Kepler tesla cards and the gtx700 line.