[ANN] cudaMiner & ccMiner CUDA based mining applications [Windows/Linux/MacOSX]

[danws6]

I tweaked my config a bit, changed to not use the cpu and different kernel for my 770's

cudaminer.exe -H 2 -d 0,1 -i 0 -l K24x16

My cpu temps are much lower now so I bumped up the gpu's to 90% power target, hitting right at 600 kh/s. I'll let this run today and see how it does.

[cbuchner1]

HAHA no not into STD's...

Wonder what I'm missing to get better performance.

play with -H 1 and -H 2 to see if it makes any difference, maybe.

[manofcolombia]

After updating my gpu won't reach its max overclock like it used to on the older config.
Its a MSI 660 ti PE and I had a +80 offset on the core clock with afterburner and it reaches 1256 mhz all the time and it has since i dialed in the overclock when I got the card. I know the card isnt going bad for sure since this only happens when I'm mining with the 12/18 version. With the same settings as before, I was mining at the consistent rate of 220 khash/s @ 1256 mhz and 92% TDP power limit. Now with 12/18, I'm mining at a rate of 291 khash/s but only at 1176/1133 mhz (it bounces between the two every second) and at a 100-110% TDP. Of course, the update was a very very nice boost but I wanna be able to run at my normal overclocking setting. Even with stock card settings I'm now running at 100+% TDP when before the update stock settings were around 75% TDP. Just seems weird and I've tried tweaking the voltages and funny enough if I up my voltage to my gpu core my clock speed drops...but if I lower the voltage the clock speed raises back up and doesnt quite reach the normal 1256 but gets close. Any ideas?

These are my mining configs

-H 1 -i 0 -C 1 -D -l K14x8

Currently getting 292-295 khash but I wanna break 300

EDIT: temps are perfectly fine hovering between 60-65C depending if I have the door open to the rest of the house and ambient temp is 23C/73F

[blackraven1425]

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.

Any more details on this? I have the 14x14 config running on my 670, and am only at ~244, even with interactive mode disabled.

[SavellM]

Wonder what I'm missing to get better performance.

play with -H 1 and -H 2 to see if it makes any difference, maybe.

Its definitely a bit better with -H 2
There is about a 30KH/s difference.

[Vbs]

if I use  -l T15x24  with good results on 780Ti with 15 SMX you should be able to use -l T14x24 on the Titan. I believe it has 14 SMX enabled, whereas the GTX 780 has 12 SMX.

Also be sure the card is not configured for double precision computation (it makes it slower for single precision work loads)

I also had that one bizarre autotune result with 1000 kHash/s, but that was too good to be true... and in fact it wasn't working for me either.

Getting 650kH/s on T15x24 on a Gigabyte 780Ti OC edition with the latest cudaMiner version.

[trell0z]

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.

Any more details on this? I have the 14x14 config running on my 670, and am only at ~244, even with interactive mode disabled.

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.

[SavellM]

OK so I've managed to push it into the 495-505KH/s mark

Using T14x32.

[cbuchner1]

If you guys could opvote the story "NVidia Kepler mining improvements now in CudaMiner - have fun!" in the litecoinmining subreddit, that would be great.

This needs some visibility. Click on meme to follow through to this story.

[cbuchner1]

OK so I've managed to push it into the 495-505KH/s mark

Using T14x32.

err... uh. Didn't know it was possible to go beyond 24. Autotune for sure cuts out at 24 at the moment.

I will make sure autotune tests more warps on this particular kernel in future versions.

[cbuchner1]

Getting 650kH/s on T15x24 on a Gigabyte 780Ti OC edition with the latest cudaMiner version.

You're way in AMD turf now.

[manofcolombia]

If you guys could opvote the story "NVidia Kepler mining improvements now in CudaMiner - have fun!" in the litecoinmining reddit, that would be great.

This needs some visibility.

Done and commented

[cbuchner1]

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.

Noted. I will see what I can do to get back to previous heat emissions.

[Number6]

Went from ~220 kHash to over 300 with same settings and a slight overclock on GTX 660ti.

Launch settings are: -H 1 -d 0 -i 0 -C 1 -l K14x16

[manofcolombia]

Went from ~220 kHash to over 300 with same settings and a slight overclock on GTX 660ti.

Launch settings are: -H 1 -d 0 -i 0 -C 1 -l K14x16

NICE! I'm at 302 now with that config, but still not getting my max overclock like I normally am

[mahowi]

Hi!
I have found this script in some other forum and it works with cpuminer. But with cudaminer, if primary pool is unreachable, it stops with

Code:

workio thread dead, waiting for workers...

and cudaminer doesn't terminate. So the script can't switch over to the next pool.
It seems cudaminer doesn't shutdown all worker threads.

This would be a bug related to stratum. Does it terminate when you try the getwork protocol?

I don't know how to test this. The only server having problems with downtimes is hashcows and it uses only stratum protocol.

Happened again:

Code:

[2013-12-18 17:34:36] stratum_recv_line failed
[2013-12-18 17:34:36] Disconnecting stratum connection...
[2013-12-18 17:34:36] Stratum connection interrupted
[2013-12-18 17:34:36] GPU #0: GeForce GT 540M, 708608 hashes, 25.97 khash/s
[2013-12-18 17:34:38] Stratum connection failed: Failed connect to stratum01.hashco.ws:8888; No error
[2013-12-18 17:34:38] ...retry after 5 seconds
[2013-12-18 17:34:45] Stratum connection failed: Failed connect to stratum01.hashco.ws:8888; No error
[2013-12-18 17:34:45] ...terminating workio thread
[2013-12-18 17:34:45] workio thread dead, waiting for workers...

Is there any way to force terminate cudaminer after some timeout?

[cbuchner1]

Is there any way to force terminate cudaminer after some timeout?

will have that in next version....

[Number6]

Went from ~220 kHash to over 300 with same settings and a slight overclock on GTX 660ti.

Launch settings are: -H 1 -d 0 -i 0 -C 1 -l K14x16

NICE! I'm at 302 now with that config, but still not getting my max overclock like I normally am

Yeah same here, I did up the power budget to 115% and it increased hash rate for like a minute, but then the voltages started bouncing around and hash rate started to go all over the board. I backed it down to 110% and it settled down again.  I think I am just going to let it run like this for a few hours to test stability before playing around anymore.

[CheatBean]

I just wanted to chime in here, I have a gtx770 that went from 220-230 to 360-370 using -l auto. Will mess around with it later to see if I can get it higher.

So great! Thanks!

[cbuchner1]

On GTX 780 (Ti) and Titan cards, one of the best launch configurartions seems to be

T number of SMX x 32

i.e. T15x32 for 780 Ti
T14x32 for Titan
T12x32 for GTX 780

Upped from 550 to 570 kHash simply by going from x24 to x32. autotune will currently not find x32 configurations.

Christian