I'm running a laptop with two 680m's and experiencing a similar issue. Autodetect was pretty lame in what it chose (it tried but I was only getting like 80kh/s) and right now I'm running two instances, one per GPU, one at K14x16, the other at K70x2 - the K14x16 one works well at around 117kh/s, the 70x2 config is a little slower at around 104kh/s

I tried running both with K14x16, however, I have a feeling I was running out of VRAM (these are the 2GB 680m's) because my drivers would crash and both cards would drop to like 5kh/s

Memory utilization at the current settings is around 87% on both cards according to HWinfo

I'm happy with around 215kh/s from a laptop with NVidia GPUs :]

That said, if anyone knows a config that may work better on the 680m, I'm all ears

Eiz,

Hey, I'm super new here too though I too made the mistake back in 2009 of thinking Bitcoin was a "fad" that would pass. I was actually "donating" the CPU power of 3 computers at the time to a prime number network...if only I had been mining instead

I think the biggest issue with 28nm vs 15nm is that of startup costs

15nm is still a very new development, and as such, only companies with the appropriate R&D budgets currently have access to it.

I'm certain that it will make it's way into the Bitcoin world soon, however, as I understand it, just as important as the mining speed would also be the mining efficiency.

This new chip process will be much more helpful in offering more computational power per KWh than it will be in producing raw GH or TH/s

Thus while network capacity (and unfortunately difficulty) will jump again, I think the bigger story will be the decreased effect on energy prices in determining margins.

This is the main reason I'm focusing mainly on Altcoins that are scrypt based... still a chance there to GPU mine whereas I totally missed out on that with Bitcoin

-Derek