Cudaminer doesn't actually use the CPU to "mine", just the sha256 stuff that can now also be done on the GPU itself. So if you want to mine with your CPU also, you need a separate miner.

You could try skyn3ts bioses (unlocks pretty much everything you'd want) http://www.overclock.net/t/1438886/official-nvidia-gtx-780-ti-owners-club (scroll down and read ^^)

Yeah, just use the 12-10 version until/if he can fix the increased temps w/o performance increase on older cards, or at least on the 580 since my result with the new version is the same ^^ I mean, if there was an equal performance increase for the heat I'd be fine with it, but absolute waste when it performs the same/slightly less with all that extra heat.

I'm guessing that should be -i 1

That's a very old version, go get the latest one and see if it works better: http://www.guru3d.com/files_details/msi_afterburner_beta_download.html

Full config is -H 1 -i 0 -m 1 -l K14x14 x64 exe, win 7, 331.93 drivers. I'm guessing you haven't overclocked your card either, that card boosts to 1280~1300mhz core, memory sits at around 3500mhz.

Ran it some more on my 580 too, heat is way crazy now, performance is the same as 12-10 build, but where it stopped at around 76C there, it now goes up to 84 and beyond :/ Worse than furmark haha. Cooled with an accelero xtreme II at 100%, so a 10C increase is quite a lot.

Oh and also auto with debug on seems to fail a ton on the 670 now, gives some 1&jhash /similar error config then starts hashing at <20khs. Without debug it's a really low config like 2x1 6x4 or similar, which also sucks.

Mines the same performance wise, although running hotter the same as yours. Major improvement on an 670 I tried on though, best previous was 230ish, now it's at 330-350! With above mentioned 14x14 conf, can't remember the previous one.

Tried it with -H 0/1 and different combinations of -C 0/1/2, 32/64bit exe ? And also disabling sli, since I don't have sli/xfire I don't know that much about the problems myself, but everywhere I've read about it people seem to say it's a bad idea when mining, no idea if it's the same or not with cudaminer though.

-H 2 with the new version gives me 10kH/s less on both x86 & x64, and also x64 is still marginally faster for me than x86.
On win 8.1 x64.

Oh ok cool! Will try to experiment a bit then.
Do you think you could change the text there to be a bit more obvious about it actually working? I mean I understand this isn't exactly something everyone will do, but should be a minor code change I guess with just some text?

Yeah it's the whole -l F / -l f (and that's even the correct kernel for my card, 580) etc thing that doesn't work, at least as far as I can understand the readme we're supposed to be able to do something like that, and it will autotune with my choosen settings, and for the kernel that I specify from the list of L/F/K/T.
Only result I get though is either "Given launch config 'x' does not validate" or the program crashes so yeah.. Otherwise I'm using the given launch config of F16x14 which I got out of normal autotune run with -D yes.

I'm quite aware of the normal autotuning, what I want to try is force it to autotune with another kernel and see if there could be any gains like you found, albeit unstable.
Problem is, with capital letters it's just "unknown", with non-capital the program just crashes, even with "f", which is the correct one for my card..

Interesting! Thx for the summary.

How do you make it autotune for an specific generation btw? Would like to experiment like blackraven here above. Am I just missing something very obvious here or is it just broken? All kinds of combinations I can think of end up in either crashes, even when using the correct Fermi config (-l f?) which is my card, or just "unknow operator"

Based on my understanding of the readme, try removing one card and then find the best combination for the other card, then put the other one in and do the same thing. Then use the appropriate launch config for each card. Try it with -i 0 -H 1 -C 0/1/2 -D. Do a few runs and see which one gives you the highest hash rate, note down the launch config (letter)number x number, then use that with the -l operator.

Edit: guess you could specify an device with the -d operator and not have to remove or disable anything

For reference, I'm getting 99% usage and 320kH/s on my 580, albeit highly overclocked but still, you should be getting quite a lot more imo.

http://forums.guru3d.com/showthread.php?t=379506 Display driver uninstaller, good for clearing everything out before a new install. Newest driver as of right now is 331.93, didn't give me anything over 82 though, it does have newer sli profiles for games though which you would benefit from I guess.
If you want to use different setups for the cards you can do that, check the readme, it explains how to with an example setup. Otherwise it'll just use the same settings for both if you don't specify anything else far as I understand it.

Strange.. guess it's the other hardware/driver ver diff. in our rigs then. Or our cards are just veery different haha.
Anyways, why the --no-autotune? Far as I can understand it, if you specify an launch config like we do with (letter)numberxnumber any kind of autotuning is disabled.

Really, what other settings? 16x16 with my config gives me only ~200kH/s

Early christmas present for all 580 (and maybe 570?) users:
-H 1 -i 0 -l F16x14 -C 2 x64 exe
Currently hashing at ~320kH/s! My 580 is running @970Mhz core / 2200Mhz mem. Mem speed doesn't affect much if at all, biggest increase is core. So if you can clock higher by lowering your memory, make a mine profile for that and another for normal games if you need the mem speed, usually though core is better there too.