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

[cbuchner1]

A crash with every kernel probanly means: update your nVidia drivers.

[1714969972]

1714969972

[1714969972]

1714969972

[1714969972]

1714969972

[1714969972]

1714969972

[1714969972]

1714969972

[cbuchner1]

For those wanting to compile the new versions on linux, once you are in the source directory:

the source code on github won't have this problem...

[ukmadorian]

A crash with every kernel probanly means: update your nVidia drivers.

Hi!

I updated my Drivers yesterday with a new release.

[cbuchner1]

I updated my Drivers yesterday with a new release.

ok then. Autotune crashing, but (some) of the kernels running when manually specified is really bizarre. At the moment I do not have a solution. Pass --no-autotune maybe and do NOT specify a -l argument. It will then fall back to a heuristical selection for the launch configuration.

By the way. according to the Google docs spreadsheet the kHash/s winner for a single nVidia equipped PC is: 8x Tesla M2050: 1252 kHash/s. Wow. I wonder what the total Wattage is.

[ukmadorian]

Thanks,

I try this option 

[robot_meh]

After having gone through the pain of compiling and configuring cudaminer for Linux I thought I should put a little how to help those with Linux machines.

My set up is default Xubuntu 12.04 straight out of the box.

You will need CUDA framework and NVIDIA graphics drivers installed. Drivers in most cases recognised and installed automatically so you should not have any problems with that. Also CUDA software is now fully supporting 12.04 version of Ubuntu, so you should not have any problems either.

For the cudaminer bit you will need to have some packages installed. In my case it was automake tools and curl library.

Here are the steps that I had to take in order to get mine working:

Note, you should get the source code from github https://github.com/cbuchner1/CudaMiner

Code:

1. sudo apt-get instsall automake
2. sudo apt-get install libcurl
3. autoreconf --install or try ./autoconf.sh
4. ./configure
5. make

If you having issues with nvcc, you need to add LD_LIBRARY_PATH to your profile:

From devtalk.nvidia.com:

In the file .bashrc in /home/username (in your home directory, google for what that is ..), add:

Code:

export LD_LIBRARY_PATH=/usr/local/cuda/lib
export PATH=$PATH:/usr/local/cuda/bin

You should be ready to launch your cudaminer.

I have Tesla C2075 and Quadro 600 both are Fermi architecture. Tesla C2075 is at device 0 location and Quadro 600 at device 1 location. Having played with some of the settings I recommend the following launch options:

Code:

./cudaminer -H 1 -d 0,1 -i 0,1 -l F56x16,F2x16 -C 2,2 -o your.stratum.server:port -O your_username.worker_name:worker_password

Here is my output:

Code:

[2013-11-20 15:30:35] Stratum detected new block
[2013-11-20 15:30:35] GPU #1: Quadro 600, 302080 hashes, 26.25 khash/s
[2013-11-20 15:30:35] GPU #0: Tesla C2075, 659456 hashes, 175.56 khash/s
[2013-11-20 15:30:50] GPU #1: Quadro 600, 400384 hashes, 26.25 khash/s
[2013-11-20 15:30:50] accepted: 419/419 (100.00%), 201.81 khash/s (yay!!!)
[2013-11-20 15:30:57] GPU #0: Tesla C2075, 4014080 hashes, 181.87 khash/s
[2013-11-20 15:30:57] accepted: 420/420 (100.00%), 208.13 khash/s (yay!!!)
[2013-11-20 15:30:59] GPU #0: Tesla C2075, 315392 hashes, 167.43 khash/s
[2013-11-20 15:30:59] accepted: 421/421 (100.00%), 193.69 khash/s (yay!!!)
[2013-11-20 15:31:01] GPU #0: Tesla C2075, 286720 hashes, 164.29 khash/s
[2013-11-20 15:31:01] accepted: 422/422 (100.00%), 190.54 khash/s (yay!!!)
[2013-11-20 15:31:05] GPU #0: Tesla C2075, 745472 hashes, 177.38 khash/s
[2013-11-20 15:31:05] accepted: 423/423 (100.00%), 203.64 khash/s (yay!!!)

Hope this helps someone.

Happy mining

[cbuchner1]

I am making steady improvements to the github repo. Added a test kernel for experimentation allowing me to try out new things. A 64bit Win binary is also on the agenda.

also: ordered a GTX 780Ti...

[HeatSurge]

also: ordered a GTX 780Ti...

Can you post benchmarks when you get it at default clocks please?  I don't trust the online google doc, and it might influence my decision to get one around xmas time vs a R9 290.

Also note that -C 1 or -C 0 might be faster than -C 2, but I'm sure you know that.  In my experience, you should autotune with -C 2 but then try -C 1 for -C 0 with the same autotune conf. for faster speeds ;-P .

[fruittool]

aaaand a new win 64bit exe

http://www.filedropper.com/cudaminer2013-11-20x64cuda-55

Say thanks to cbuchner1, he made this all possible....

Me, well i just pressed a few buttons.

Have fun.

[cbuchner1]

I posted a small update for the 2013-11-20 version with a 32+64 bit binary. The 64 bit binary will do the SHA256 part on the CPU more efficiently, which lowers the CPU utilization. I also added a few checks if CUDA was opened successfully and if the driver meets the version requirements.

Let me know if you see any mining speed degradation in the 64 bit build with specific kernels or hardware.

fruittool, thanks for providing the 64 bit versions so far. I guess you can sit back and relax now.

Christian

[mhps]

I posted a small update for the 2013-11-20 version with a 32+64 bit binary. The 64 bit binary will do the SHA256 part on the CPU more efficiently, which lowers the CPU utilization. I also added a few checks if CUDA was opened successfully and if the driver meets the version requirements.

Let me know if you see any mining speed degradation in the 64 bit build with specific kernels or hardware.

My tesla c2075 is doing 181 kh/s compared with 165 a month ago with -l F14x16 -i 0 -C 2 (power 175W temp 90C)

Very nice improvement 

[Number6]

First of all, thanks for keeping this updated!

I am getting ~220 kHash/sec on a 660Ti with launch settings of:

Code:

-d 0 -i 0 -C 1 -l K14x8

with the 32 bit client.

Using the 64 bit client, and the same launch settings as above, it drops down to 180 kHash/sec.

Other system info:

Windows 7 64 bit
Nvidia 327.23 drivers

[cbuchner1]

First of all, thanks for keeping this updated!

I am getting ~220 kHash/sec on a 660Ti with launch settings of:

Code:

-d 0 -i 0 -C 1 -l K14x8

with the 32 bit client.

Using the 64 bit client, and the same launch settings as above, it drops down to 180 kHash/sec.

Other system info:

Windows 7 64 bit
Nvidia 327.23 drivers

what about fruittool's build? Same problem there?

[Number6]

Yep, just downloaded fruittool's version as well to compare. I used the exact same launch settings as I posted above, and it is also about 40 kHash lower, or ~ 180 kHash/sec in my case. I also cross checked what the console was reporting with the statistics on my pool, and the results were consistent; lower with the 64 bit clients.

Back to using the 32 bit version for now and 220 kHash/sec.

[Schleicher]

On my system:
134kh/s with 64bit
132kh/s with 32bit

That's a GTX 660 and a AMD Phenom II X4 910e running under Win 7
with -i 0 -C 2 -l K80x2

[atomton]

Hey guys,

I've managed to successfully compile cudaminer on OS X 10.9 – if anyone wants to give it a try, here's how I did it: http://atomton.roon.io/compiling-cudaminer-on-os-x-10-9.

I get around 38 kh/s compared to 23 kh/s with cgminer on a GT 650M when mining Litecoins. Would be cool if someone else has success and posts some hashrates (and/or feedback).

[fruittool]

Yep, just downloaded fruittool's version as well to compare. I used the exact same launch settings as I posted above, and it is also about 40 kHash lower, or ~ 180 kHash/sec in my case. I also cross checked what the console was reporting with the statistics on my pool, and the results were consistent; lower with the 64 bit clients.

Back to using the 32 bit version for now and 220 kHash/sec.

Did you try the -H 1 switch? Assuming you have a dual core or more. The 64 bit build seems to use more cpu and may benefit from it more or less depending on cpu speed.

[Spoetnik]

every question i have asked on this topic has been pretty well ignored..
i made a 64bit build many months ago before fruittool posted his i believe but i used cuda 5.0 not 5.5
i've mentioned this but i also noticed before chubner1 you said it did not help speed (which it did of course on Fermi i tested)
i went on since July making tons of other features and fix etc and even shared it with some people around the scene they love it.
but if i comment here i get ignored and or argued with lol
So i keep my version to myself
Mine has a color coded logging api.. no black / white

i fixed the stratum freezing issue on my builds and it didn't happen for a couple months..
then you recently posted out of no where its a problem (even though you wouldn't acknowledge it was before) that you fixed it..
so i went and compared every line of your source with mine with beyond compare and duplicated your changes
and guess what ? freezing galore.. it happens regularly now lol
compared to not at all for 2/3 months..

other FIX's and feature improvements also

i also made a MFC gui mod using your cuda code but its kinda old now and would be windows only.
which features galore such as tab or treeview mode dedicated file and or console output.. config section etc
its no complete yet though and since its windows only i wasn't sure if anyone cares.. no one seemed to when i mentioned here before (ignored again lol)

here is a screenshot of my current 64bit build on Win 7 using cuda 5.0
notice the scantime info and how i mentioned already i fixed the major bug / issue
i mentioned already about how the scan time doesn't do shit unless your not using Getwork or not using Stratum (and was of course once again ignored lol)



anyway i'm buying a Radeon 7950 so im not sure i'll care as much now hahah

[HeatSurge]

I posted a small update for the 2013-11-20 version with a 32+64 bit binary. The 64 bit binary will do the SHA256 part on the CPU more efficiently, which lowers the CPU utilization. I also added a few checks if CUDA was opened successfully and if the driver meets the version requirements.

Let me know if you see any mining speed degradation in the 64 bit build with specific kernels or hardware.

fruittool, thanks for providing the 64 bit versions so far. I guess you can sit back and relax now.

Christian

My rate increased from 275 to 283 with the same everything from here:

https://bitcointalk.org/index.php?topic=167229.msg3648834#msg3648834

My older K15x16 -C 0 config (when fermi was crashing a couple of builds ago) dropped from 270 to 267, but I'm not using that one anymore so I don't care.

This keeps getting better...  300 soon?  :-P  Don't forget to post benchies from your GTX 780 Ti when you get it plz!

[mrm0]

[...]
i made a 64bit build many months ago before fruittool posted his i believe but i used cuda 5.0 not 5.5
i've mentioned this but i also noticed before chubner1 you said it did not help speed (which it did of course on Fermi i tested)
i went on since July making tons of other features and fix etc and even shared it with some people around the scene they love it.
but if i comment here i get ignored and or argued with lol
So i keep my version to myself
[...]
i fixed the stratum freezing issue on my builds and it didn't happen for a couple months..
[...]
other FIX's and feature improvements also
[...]
(and so on, and so forth)

Well, so you took a working Open Source code, made some improvements, and *even* shared with a couple of friends. Do I read that right?

How very generous of you...
(in case you didn't notice, this was sarcastic)

Please tell, why didn't you make a fork on github instead?
Oh, and that screenshot of yours.. I do not see even a slight mention of the original author(s). Not nice.

Anyway, the recent 'official' version from cbuchner1's github repo compiles and works for me just fine.
Dual GTX480 setup. Around 440 kh/s combined.