https://bt.irlbtc.com/view/377174.msg4040564#msg4040564

Just posting the following for a newbie from the above thread:

i have error at cudaminer:
please help.

Well thats what I was saying in a post earlier, since the update of 12/18 my gpu is now taking 103%+ power when it used to sit nicely at 95%
And if i up my voltages to where they normally are for stable gaming my core speed drops and so does my hashrate. So really right now these settings will instantly crash with any game or anything intense. Thats because my voltage is so low because I wanna keep my clock speed high for mining, even though it doesnt get nearly as high as it used to. As for lifespan, this is a MSI Power Edition card so I'm confident with the integrity however I don't understand how my power usage shot up like by a full 10% after 12/18 with the exact same settings I was using in 12/10.

EDIT: Also, what pool are you using because I was seeing the same results all yesterday. I was getting just chunks of stales in 5-10 when normally I'm at a 99% valid yesterday was like at 96%.

can't use x4. Stick to x3 launch configs

No luck with my 580 and the latest version, its basically cooking itself with no performance increase.
gtx660 got a nice performance bump  

all sha256 and scrypt types

but at what hash rate? for 780gtx

I don't think it's a great idea.

I'm confused about your performance.  I now have my 690 working well.  It's at 85degC on one half and 77degC on the other half doing 550 kh/sec (with no display attached).

That's with kernel config -l K8x32,K8x32

What temperature are you seeing?

If you want to edit the source, go into kepler_kernel.cu and look for the line that says
Sleep(1);

and change it to Sleep(10);

and then recompile.  See how much of a temperature reduction that gives you vs performance drop and play from there.

But, as I said - your performance seems low.  You might have a ventilation problem that's causing some thermal shutdown?

  -Dave

Not yet.  I'm finally managing to reproduce your thermal overload problem on my own setup with 2x GTX690s:

| 82%   90C  N/A     N/A  N/A |   1087MiB  2047MiB |     N/A      Default |
| 57%   78C  N/A     N/A  N/A |   1087MiB  2047MiB |     N/A      Default |
| 57%   79C  N/A     N/A  N/A |   1087MiB  2047MiB |     N/A      Default |
| 54%   75C  N/A     N/A  N/A |   1087MiB  2047MiB |     N/A      Default |

Toasty.  That 90C isn't good unless planning on making tea on your computer.

My kernel is going to make your display laggy even in interactive mode, unfortunately.  The only thing I can think of to try to reduce both power and lagginess without changing the code is to try -l K2x32,K2x32 or something similar.  Have you given that a shot?  It should reduce the duration of time that the kernel runs and increase the relative amount of time interactive mode spends telling the GPU to not do mining.  Could you let me know how that works?

I'm tied up for a while, but I'll see if now that I have my 690 running I can figure out any efficiency gains for it.  Don't hold your breath, though:  The 690 is pretty similar to the Grid K2 that I was optimizing for before.  I think there are gains to be had for GF110 devices, but maybe not GF104.

  -Dave

My 770GTX produces 330kH/s at  -i 0 -C 2 -m 1 -H 2 -l K16x16 (shitty CPU does not like -H 1 )

However for 24h operation I had to lower voltage and underclock to get temps stabilize to 79 degrees of Celsius.(I will optimize this later) With underclock I get 300kH/s.