Ubuntu 11.04 Natty 64bit Headless Bitcoin mining with ATI 11.5 binary driver

[dukejer]

Hi Everyone,

I was able to repurpose an existing computer now running headless in text console mode only with two new AMD Radeon HD5830's cards over clocked to 875MHz.  The GPUs are running at a stable 86 degrees Celsius and average 519 Mhash/s between the two HD5830's.

*** Warning this may cause irreparable damage to your computer system, harm yourself and your family, burn down your house, neighborhood and city, generate untold amounts of carbon dioxide at your electric company and harm the polar bears and baby seals in the Arctic Circle.  I am NOT responsible if you try this or for any errors that may exist within this post.

Load a fresh Ubuntu Natty 11.04 64-bit Desktop with the latest updates and log into system with a user that has sudo permissions.
   sudo apt-get remove nvidia-common
   sudo apt-get install libqtgui4

Load python and other development tools
   cd ~
   sudo apt-get install python-setuptools python-numpy subversion g++ libboost-all-dev

Download and install ATI Driver 11.5 for Linux 64bit.
   cd ~
   wget http://www2.ati.com/drivers/linux/ati-driver-installer-11-5-x86.x86_64.run
   sudo sh ati-driver-installer-11-5-x86.x86_64.run --buildpkg Ubuntu/natty
   sudo dpkg -i *.deb
   sudo apt-get -f install
   sudo aticonfig -f --initial --adapter=all
   sudo reboot

Verify that the ATI Driver is setup and running
   cd ~
   DISPLAY=:0 sudo fglrxinfo

Download and install bitcoin
   cd ~
   wget http://downloads.sourceforge.net/project/bitcoin/Bitcoin/bitcoin-0.3.21/bitcoin-0.3.21-linux.tar.gz
   tar xzvf bitcoin-0.3.21-linux.tar.gz
   chmod +x bitcoin-0.3.21/bin/64/bitcoin*
   mkdir -p ~/.bitcoin
   echo "rpcuser=user" >> ~/.bitcoin/bitcoin.conf
   echo "rpcpassword=password" >> ~/.bitcoin/bitcoin.conf

Install python-jsonrpc
   cd ~
   svn checkout http://svn.json-rpc.org/trunk/python-jsonrpc
   cd python-jsonrpc/
   sudo python setup.py install

Download and install AMD APP SDK 2.4 for Linux 64bit.
   cd ~
   wget http://download2-developer.amd.com/amd/APPSDK/AMD-APP-SDK-v2.4-lnx64.tgz
   tar xvzf AMD-APP-SDK-v2.4-lnx64.tgz
   echo export AMDAPPSDKROOT=${HOME}/AMD-APP-SDK-v2.4-lnx64/ >> ~/.bashrc
   echo export AMDAPPSDKSAMPLESROOT=${HOME}/AMD-APP-SDK-v2.4-lnx64/ >> ~/.bashrc
   echo 'export LD_LIBRARY_PATH=${AMDAPPSDKROOT}lib/x86_64:${LD_LIBRARY_PATH}' >> ~/.bashrc
   source ~/.bashrc
   cd /
   sudo tar xfz $AMDAPPSDKROOT/icd-registration.tgz

Download, Compile and Install pyopencl-0.92
   cd ~
   wget http://pypi.python.org/packages/source/p/pyopencl/pyopencl-0.92.tar.gz
   tar xzvf pyopencl-0.92.tar.gz
   cd pyopencl-0.92
   ./configure.py --cl-inc-dir=${AMDAPPSDKROOT}include --cl-lib-dir=${AMDAPPSDKROOT}lib/x86_64
   make
   sudo make install

Download and Install Phoenix Miner 1.48
   wget http://svn3.xp-dev.com/svn/phoenix-miner/files/phoenix-1.48.tar.bz2
   tar xvf phoenix*.bz2

Verify that OpenCL is setup and running
   cd ~
   cd AMD-APP-SDK-v2.4-lnx64/bin/x86_64
   ./clinfo

Create a startminer script using code from below.  Make sure to substitute the correct home directory path, miner pool server, miner user and miner password.
   cd ~
   sudo touch /usr/local/bin/startminer.sh
   sudo chmod 755 /usr/local/bin/startminer.sh
   sudo gedit /usr/local/bin/startminer.sh

      # ${1} is used as a variable for the username, password and for the gpu device number.  Ex. mineruser0, minerpass0, Device=0 or mineruser1, minerpass1, Device=1
      HOMEDIR=/home/user
      MINERSERV=btcmine.com:8332
      MINERUSER=mineruser@miner${1}
      MINERPASS=minerpass${1}

      export AMDAPPSDKROOT=${HOMEDIR}/AMD-APP-SDK-v2.4-lnx64/
      export AMDAPPSDKSAMPLESROOT=${HOMEDIR}/AMD-APP-SDK-v2.4-lnx64/
      export LD_LIBRARY_PATH=${AMDAPPSDKROOT}lib/x86_64:${LD_LIBRARY_PATH}

      #Overclock GPU to 875Mhz
      DISPLAY=:0 aticonfig --od-enable --adapter=all
      DISPLAY=:0 aticonfig --od-setclocks=875,1000 --adapter=${1}
      cd ${HOMEDIR}/phoenix-1.48
      echo "Startming Miner: ${1}"
      ${HOMEDIR}/phoenix-1.48/phoenix.py -u http://${MINERUSER}:${MINERPASS}@${MINERSERV} -k phatk VECTORS BFI_INT AGGRESSION=12 DEVICE=${1}

Setup Headless Bitcoin Mining
 *** Warning *** This will stop your computer from booting a graphical desktop and allow only text console or remote ssh access into the mining server.
 If you rely on a GUI for administration you may want to rethink this.
   sudo apt-get install openssh-server
   sudo apt-get install screen
   sudo mv /etc/init/gdm.conf /etc/init/gdm.org
   sudo gedit /etc/init/startx.conf
        description     "Start X Server for btc mining"
        start on runlevel [2345]
        stop on runlevel [!2345]
        kill timeout 30
        script
           exec /usr/bin/X 2>&1
        end script

   sudo gedit /etc/init/btcminer_0.conf
        description     "Start BTC Mining"
        start on runlevel [2345]
        stop on runlevel [!2345]
        kill timeout 30
        script
           LOGINUSER=ChangeToMyLoginUser
          #Wait 30 seconds to make sure X is started.
          sleep 30
          exec /usr/bin/screen -dmS gpu0 su -c '/usr/local/bin/startminer.sh 0' ${LOGINUSER}
        end script

   sudo gedit /etc/init/btcminer_1.conf
        description     "Start BTC Mining"
        start on runlevel [2345]
        stop on runlevel [!2345]
        kill timeout 30
        script
        LOGINUSER=ChangeToMyLoginUser
        #Wait 35 seconds to make sure X is started.
        sleep 35
           exec /usr/bin/screen -dmS gpu1 su -c '/usr/local/bin/startminer.sh 1' ${LOGINUSER}
        end script

How to access the miner server
   ssh into the mining server as your normal login user.
   sudo screen -r gpu0 or sudo screen -r gpu1

To disconnect from screen Control-A + Control-D.

How to check the GPU temperatures

   ssh into mining server as your normal login user.
   DISPLAY=:0 sudo aticonfig --odgt --adapter=all

How to monitor your screens on an Xterminal on the main Xwindows console on boot up.

   sudo gedit /etc/init/btcmonitor_0.conf
        description     "Start BTC Monitor 0"
        start on runlevel [2345]
        stop on runlevel [!2345]
        kill timeout 30
        script
          LOGINUSER=ChangeToMyLoginUser
          export DISPLAY=:0
          #Wait 40 seconds to make sure X is started.
          sleep 40
          exec  /usr/bin/xterm -geometry 80x40+0+0 -e "/usr/bin/screen -d -r gpu0"
        end script

sudo gedit /etc/init/btcmonitor_1.conf
        description     "Start BTC Monitor 1"
        start on runlevel [2345]
        stop on runlevel [!2345]
        kill timeout 30
        script
          LOGINUSER=ChangeToMyLoginUser
          export DISPLAY=:0
          #Wait 45 seconds to make sure X is started.
          sleep 45
          exec  /usr/bin/xterm -geometry 80x40+500+0 -e "/usr/bin/screen -d -r gpu1"
        end script
   


I have compiled this with the help from many sources from around the internet including.

http://sectio-aurea.blogspot.com/2011/05/bitcoin-mining-on-ubuntu-1104-natty.html
http://forum.bitcoin.org/?topic=7514.0

Along with my own 30 years of computer experience and 24 years of Unix and Linux.

If you find this post helpful feel free to donate to address: 19oQLN2v1auiGqBDFfxSDUfu8Ax6ooT6xs


- Dukejer

      

[leepfrog]

This is an extremely helpful tutorial and I am looking forward to use it when configuring my rig at the beginning of next week.
Thanks for your effort!

[Syke]

I have a much simpler setup, but I am mining headless successfully (no monitor, keyboard, or mouse plugged in at all). What I can't do is aticonfig. It complains about X not running. Do I have to hack up the /etc/init scripts to get X running without a monitor plugged in?

[elrock]

I have a much simpler setup, but I am mining headless successfully (no monitor, keyboard, or mouse plugged in at all). What I can't do is aticonfig. It complains about X not running. Do I have to hack up the /etc/init scripts to get X running without a monitor plugged in?

These three lines worked for me.

[dukejer]

I have a much simpler setup, but I am mining headless successfully (no monitor, keyboard, or mouse plugged in at all). What I can't do is aticonfig. It complains about X not running. Do I have to hack up the /etc/init scripts to get X running without a monitor plugged in?

Hi Syke,

You are able to mine without running X Windows?   I was not able to run atiConfig or mining without X running.  If you want to start X you should be able to do it through a small script in /etc/init or in one of the /etc/rc scripts.  You will need to set the DISPLAY=:0 command if you are remote and also you may need to run sudo aticonfig since X will be owned by root once it starts.

- Dukejer

If you find this post helpful feel free to donate to address: 19oQLN2v1auiGqBDFfxSDUfu8Ax6ooT6xs

[Syke]

Well I guess my problem isn't due to running headless. When I boot with a monitor and see X (Ubuntu Desktop 10.10) up and running, I still can't aticonfig over ssh. The three lines didn't help either.

[dukejer]

Well I guess my problem isn't due to running headless. When I boot with a monitor and see X (Ubuntu Desktop 10.10) up and running, I still can't aticonfig over ssh. The three lines didn't help either.

Hi Syke,

Did you try this from ssh?  

DISPLAY=:0 sudo aticonfig --odgt --adapter=all

This is what I use.  You may need to be logged into X for this to work.  If you are still running the desktop try and have gdm autologin your user under the console.  You can set up autologin for Ubuntu under System -> Administration -> Login Screen.



- Dukejer

If you find this post helpful feel free to donate to address: 19oQLN2v1auiGqBDFfxSDUfu8Ax6ooT6xs

[lexa500]

Thanks for the manual!
but i have problem with pyopencl. it looks like i have missed some links to the sdk libraries

Code:

/usr/bin/ld: cannot find -lOpenCL
collect2: ld returned 1 exit status
error: command 'g++' failed with exit status 1
make: *** [all] Error 1

[dukejer]

Thanks for the manual!
but i have problem with pyopencl. it looks like i have missed some links to the sdk libraries

Code:

/usr/bin/ld: cannot find -lOpenCL
collect2: ld returned 1 exit status
error: command 'g++' failed with exit status 1
make: *** [all] Error 1

Hi lexa500,

Did you make sure your LD_LIBRARY_PATH is set correctly?  You can check it by echo $LD_LIBRARY_PATH.

Did you see any errors while compiling pyopencl?  You can rerun make and see if any errors show.

Also are you running 32-bit or 64-bit?  The links in my post were for 64-bit.

- Dukejer

If you find this post helpful feel free to donate to address: 19oQLN2v1auiGqBDFfxSDUfu8Ax6ooT6xs

[lexa500]

thanks! i`ve found the missing path

[dukejer]

I made one minor change in the overclocking.  I forgot to show how to enable overclocking first in aticonf.
Here are the changes.  If anyone sees any other errors please let me know and I will update the main post.

      #Overclock GPU to 875Mhz
      DISPLAY=:0 aticonfig --od-enable --adapter=all
      DISPLAY=:0 aticonfig --od-setclocks=875,1000 --adapter=${1}

Thank you,

-DukeJer

If you find this post helpful feel free to donate to address: 19oQLN2v1auiGqBDFfxSDUfu8Ax6ooT6xs

[leepfrog]

thanks! i`ve found the missing path

Could you please explain how? I am still stuck at this error.
The path seems to be fine:

Code:

<user>@miner:~$ echo $LD_LIBRARY_PATH
/home/<user>/AMD-APP-SDK-v2.4-lnx64/lib/x86_64:

Also the folder does contain files:

Code:

<user>@miner:~$ ls /home/<user>/AMD-APP-SDK-v2.4-lnx64/lib/x86_64
libamdocl64.so  libGLEW.so  libglut.so  libOpenCL.so  libOpenCL.so.1

However I am getting the exact same error as you.

Thanks in advance!

[dukejer]

thanks! i`ve found the missing path

Could you please explain how? I am still stuck at this error.
The path seems to be fine:

Code:

<user>@miner:~$ echo $LD_LIBRARY_PATH
/home/<user>/AMD-APP-SDK-v2.4-lnx64/lib/x86_64:

Also the folder does contain files:

Code:

<user>@miner:~$ ls /home/<user>/AMD-APP-SDK-v2.4-lnx64/lib/x86_64
libamdocl64.so  libGLEW.so  libglut.so  libOpenCL.so  libOpenCL.so.1

However I am getting the exact same error as you.

Thanks in advance!

Hi Leepfrog,

We found some cut and paste errors in my original script when I was writing the post.  Some of the lines with the "--" dash dash symbol were being substituted with a long hyphen instead.  You will probably have to run the following lines below to correct it.

  cd ~
  cd pyopencl-0.92
  rm ./siteconf.py
 ./configure.py --cl-inc-dir=${AMDAPPSDKROOT}include --cl-lib-dir=${AMDAPPSDKROOT}lib/x86_64
   make
   sudo make install

-Dukejer

If you find this post helpful feel free to donate to address: 19oQLN2v1auiGqBDFfxSDUfu8Ax6ooT6xs

[leepfrog]

Hi dukejer,

thanks for your fast response. Yes I've found those errors (wanted to post about it but they were already fixed).
However the error persisted.
I was able to solve it via this command:

Code:

./configure.py --cl-inc-dir=/home/<user>/AMD-APP-SDK-v2.4-lnx64/include/ --cl-lib-dir=/home/<user>/AMD-APP-SDK-v2.4-lnx64/lib/x86_64

So I suppose that one of the other variables must have caused problems... anyhow - working now!

Just a quick question: Is there a quick to stop or restart the miners once they have been stopped? (besides from going into every screen to ctrl+c and reboot the server to restart again)?

[dukejer]

Hi dukejer,

thanks for your fast response. Yes I've found those errors (wanted to post about it but they were already fixed).
However the error persisted.
I was able to solve it via this command:

Code:

./configure.py --cl-inc-dir=/home/<user>/AMD-APP-SDK-v2.4-lnx64/include/ --cl-lib-dir=/home/<user>/AMD-APP-SDK-v2.4-lnx64/lib/x86_64

So I suppose that one of the other variables must have caused problems... anyhow - working now!

Just a quick question: Is there a quick to stop or restart the miners once they have been stopped? (besides from going into every screen to ctrl+c and reboot the server to restart again)?

Hi Leepfrog,

You should be able to restart either of the GPUs using the service command for upstart.

sudo service btcminer_0 start
sudo service btcminer_1 start for the 2nd gpu.

- Dukejer

If you find this post helpful feel free to donate to address: 19oQLN2v1auiGqBDFfxSDUfu8Ax6ooT6xs

[Zamicol]

Good guide!  I stumbled onto this while looking for screen session help.  I've been scavenging the internet trying to figure this out:

Is there any way to make a simple, purely miner (no overclock) startup script that puts each miner into its own screen?  I've been experimenting with something like:

Code:

#!/bin/sh
screen -d -m -S "gpu0"
screen -d -m -S "gpu1"
screen -p gpu0 -X python pheonix.py -blah -blah
screen -p gpu1 -X python pheonix.py -blah -blah

Thanks!

[dukejer]

Good guide!  I stumbled onto this while looking for screen session help.  I've been scavenging the internet trying to figure this out:

Is there any way to make a simple, purely miner (no overclock) startup script that puts each miner into its own screen?  I've been experimenting with something like:

Code:

#!/bin/sh
screen -d -m -S "gpu0"
screen -d -m -S "gpu1"
screen -p gpu0 -X python pheonix.py -blah -blah
screen -p gpu1 -X python pheonix.py -blah -blah

Thanks!

Hi Zamicol,

I did not have much luck running screen as a normal user since the tty devices are created on demand now with udev even though I put my normal user in the tty group.  I normally run screen using sudo so that the super user will create the devices and then I can connect to the screen instance.  The only problem is that when I want to reconnect I also have to do a sudo screen.   The /usr/local/bin/startminer.sh script  I created in the top post can be ran without overclocking.  You can just leave out the aticonfig lines from the script that are used for overclocking.

     DISPLAY=:0 aticonfig --od-enable --adapter=all
      DISPLAY=:0 aticonfig --od-setclocks=875,1000 --adapter=${1}

You should be able to start the /usr/local/bin/startminer.sh script using the following command:

sudo /usr/bin/screen -dmS gpu0 su -c '/usr/local/bin/startminer.sh 0' YourLogin
or
sudo /usr/bin/screen -dmS gpu0 su -c '/usr/local/bin/startminer.sh 1' YourLogin
for the 2nd gpu.

- DukeJer

If you find this post helpful feel free to donate to address: 19oQLN2v1auiGqBDFfxSDUfu8Ax6ooT6xs

[dukejer]

Thank you for the donation whoever sent it.  It is nice to see a bitcoin or two from time to time.

-Dukejer

[Basiley]

jeez, just stay with Debian sid[this time, codenamed "Wheezy"], if you need power of Debian w/o Ubuntu bugs.
stay away from crippled forks such as "xbuntu" crap, IMO.

[dukejer]

jeez, just stay with Debian sid[this time, codenamed "Wheezy"], if you need power of Debian w/o Ubuntu bugs.
stay away from crippled forks such as "xbuntu" crap, IMO.

Hi Basiley,

  I have used every Linux distribution from Yggdrasil, Slackware, Caldera, Redhat, Suse, Gento, Debian, etc.  Ubuntu works fine for me.  Seems polished just enough compared to other distributions I have used in the past.  I think with a little debugging I could fix the udev issues with probably writing a udev script or some other fix but since I already have my rig autostarting in a headless configuration I have not delved into it to much.  Also taking down my rig for a bit feels like I am waisting bitcoins.  

- Dukejer

If you find this post helpful feel free to donate to address: 19oQLN2v1auiGqBDFfxSDUfu8Ax6ooT6xs

[leepfrog]

Hi Leepfrog,

You should be able to restart either of the GPUs using the service command for upstart.

sudo service btcminer_0 start
sudo service btcminer_1 start for the 2nd gpu.

- Dukejer

Thanks for the response - unfortunately this does not work. I believe the reason for that is that "service" only searches in /etc/init.d/.
Do you know any possibility to work around that?

[dukejer]

Hi Leepfrog,

You should be able to restart either of the GPUs using the service command for upstart.

sudo service btcminer_0 start
sudo service btcminer_1 start for the 2nd gpu.

- Dukejer

Thanks for the response - unfortunately this does not work. I believe the reason for that is that "service" only searches in /etc/init.d/.
Do you know any possibility to work around that?

Hi Leepfrog,

Are your receiving any errors when you run sudo service btcminer_0 start ?

service searches /etc/init for btcminer_0.  Did you create a btcminer_0.conf file in /etc/init?

Thank you,

- Dukejer

If you find this post helpful feel free to donate to address: 19oQLN2v1auiGqBDFfxSDUfu8Ax6ooT6xs

[fasti]

Just one possible thing missing, ability to underclock your memory by atleast 60% of the default ( even down to 300 if possible ). This will decrease graphic cards temperature a lot, giving you ability to overclock core more or just enjoy quieter fans.

[dukejer]

how will the procedure change if i have just installed a secondary graphical card same/different model? How do I install driver and change configuration!? Thanks

Hi hipaulshi,

You should be able to just rerun aticonfig to find the new card and then add another btcminer_X.conf file in /etc/init with the correct device number for the card in the btcminer_X.conf

The aticonfig line should be
DISPLAY=:0 sudo aticonfig --initial -f --adapter=all

- Dukejer

If you find this post helpful feel free to donate to address: 19oQLN2v1auiGqBDFfxSDUfu8Ax6ooT6xs

[dukejer]

Just one possible thing missing, ability to underclock your memory by atleast 60% of the default ( even down to 300 if possible ). This will decrease graphic cards temperature a lot, giving you ability to overclock core more or just enjoy quieter fans.

Hi fasti,

I have a new startminer script and new btcminer_X.conf scripts that can over
clock and underclock per card that hopefully I can upload in a couple of days.

- Dukejer

If you find this post helpful feel free to donate to address: 19oQLN2v1auiGqBDFfxSDUfu8Ax6ooT6xs

[coinjedi]

I have followed the instructions line by line (except I am using poclmb instead of pheonix), but I am running into a lot of issues.

1) If I do everything as written I think I am running into network issues (ssh doesn't response). How can I fall back to command line?
If I just move gdm.conf and not use startx.conf, an ubuntu logo appears with four blinking dots and then a blank screen. If I use startx.conf, x-server takes over the screen. I can't make it boot into the command line.

2) If I don't move gdm.conf, Ubuntu starts with the gui and network connection, but "screen -ls" lists nothing. How can I troubleshoot my "btcminer_0.conf"?

Thanks in advance.

[melanarchy]

So thanks to this guide I got my 6870 set up with Natty. My box is headless (2 wires plugged in, power and ethernet) so this is the perfect guide for me.

A few things I've done.
1) Get AMDOverdriveCtrl - although this is a GUI app, you can run a batch job that will let you override the minimum memory clock and allow you to underclock your memory (I haven't yet figured out a config that allows you to override the maximum gpu clock).
Here is the config for lower memory

<OVERDRIVE_PROFILE>
  <PERFORMANCE_LEVEL level="2" gpu="100000" mem="30000" voltage="1088"/>
  <PERFORMANCE_LEVEL level="1" gpu="55000" mem="30000" voltage="1038"/>
  <PERFORMANCE_LEVEL level="0" gpu="15700" mem="30000" voltage="1000"/>
  <FAN_SETTING percentage="55"/>
  <FAN_CTRL enabled="yes"/>
</OVERDRIVE_PROFILE>

Save that to a file and then run AMDOverdriveCtrl -b filename

Now using aticonfig --odgc you should see a lower memory limit. You can then do sudo aticonfig --adapter=0 --od-setclocks=1000,400 (or whatever you want) to under clock the memory.

My 6870's max stock overclock is 1000 and i haven't figured out how to get AMDOverdriveCtrl to accept a config that goes above that yet.  Once you've found what works for you don't forget to save the clocks onto your gpu so they persist through reboots with aticonfig --odcc --adapter=all.

2) Use the phatk kernel (actually I noticed the guide specifies this, I just missed it the first round through)
I've also found that with the version of the SDK outlined here the phatk kernel is better than poclbm with this execution line I get 299+ Mhash/sec out of my 6870:
 ./phoenix.py -u http://XXXXX -k phatk DEVICE=0 BFI_INT VECTORS FASTLOOP=false AGGRESSION=12 WORKSIZE=128 PLATFORM=0
 

I don't use the autostart scripts, mainly because I've been futzing with values, but I launch a screen and keep phoenix open in it, and a script to look at gpu temp/current balance in another virtual window.

I'm happy to answer questions about this if people need help.
_{1aLV7cjRXE89KoychBaGUP8YGm5ZRVNAw - If my info is helpful donations are always welcome.}

[dukejer]

I have followed the instructions line by line (except I am using poclmb instead of pheonix), but I am running into a lot of issues.

1) If I do everything as written I think I am running into network issues (ssh doesn't response). How can I fall back to command line?
If I just move gdm.conf and not use startx.conf, an ubuntu logo appears with four blinking dots and then a blank screen. If I use startx.conf, x-server takes over the screen. I can't make it boot into the command line.

2) If I don't move gdm.conf, Ubuntu starts with the gui and network connection, but "screen -ls" lists nothing. How can I troubleshoot my "btcminer_0.conf"?

Thanks in advance.

Hi coinjedi,

Did you load openssh-server?

To get to a terminal screen from the main console press control + alt + f2.  From there you should be able to login a console and put gdm.conf back into place but you need to rename startx.conf to something else without the .conf at the end first.  Also you will not be able to use the btcminer files in /etc/init but you should be able to call startminers.sh from a regular terminal within gnome.

-Dukejer

If you find this post helpful feel free to donate to address: 19oQLN2v1auiGqBDFfxSDUfu8Ax6ooT6xs

[leepfrog]

Hi Leepfrog,

Are your receiving any errors when you run sudo service btcminer_0 start ?

service searches /etc/init for btcminer_0.  Did you create a btcminer_0.conf file in /etc/init?

Thank you,

- Dukejer

Sorry, seems that I was a bit tired when trying it the first time, it works now.

BTW: I've added munin to monitor hashrate, load and temperatures as well as a script which automatically sets fan speed to max at a certain temperature threshold. If the critical threshold is passed it automatically stops the miner for the gpu running to hot. It might be dirty coding as I do not know my way around bash very well but it works for me.

If I find the time this weekend I'll describe the scripts needed

[dukejer]

So thanks to this guide I got my 6870 set up with Natty. My box is headless (2 wires plugged in, power and ethernet) so this is the perfect guide for me.

A few things I've done.
1) Get AMDOverdriveCtrl - although this is a GUI app, you can run a batch job that will let you override the minimum memory clock and allow you to underclock your memory (I haven't yet figured out a config that allows you to override the maximum gpu clock).
Here is the config for lower memory

<OVERDRIVE_PROFILE>
  <PERFORMANCE_LEVEL level="2" gpu="100000" mem="30000" voltage="1088"/>
  <PERFORMANCE_LEVEL level="1" gpu="55000" mem="30000" voltage="1038"/>
  <PERFORMANCE_LEVEL level="0" gpu="15700" mem="30000" voltage="1000"/>
  <FAN_SETTING percentage="55"/>
  <FAN_CTRL enabled="yes"/>
</OVERDRIVE_PROFILE>

Save that to a file and then run AMDOverdriveCtrl -b filename

Now using aticonfig --odgc you should see a lower memory limit. You can then do sudo aticonfig --adapter=0 --od-setclocks=1000,400 (or whatever you want) to under clock the memory.

My 6870's max stock overclock is 1000 and i haven't figured out how to get AMDOverdriveCtrl to accept a config that goes above that yet.  Once you've found what works for you don't forget to save the clocks onto your gpu so they persist through reboots with aticonfig --odcc --adapter=all.

2) Use the phatk kernel (actually I noticed the guide specifies this, I just missed it the first round through)
I've also found that with the version of the SDK outlined here the phatk kernel is better than poclbm with this execution line I get 299+ Mhash/sec out of my 6870:
 ./phoenix.py -u http://XXXXX -k phatk DEVICE=0 BFI_INT VECTORS FASTLOOP=false AGGRESSION=12 WORKSIZE=128 PLATFORM=0
,

Hi melanarchy,

Thanks for the suggestions.  I will try them out on a 5870 server I have available to me.

- Dukejer

[melanarchy]

No problem, thanks for the guide. If I discover anything else I'll come post. Really want to figure out how to override my max clock and push up a few more mhz so I can break 299.6 Mhash/sec barrier and get the 300 I know I can.

[coinjedi]

Hi coinjedi,

Did you load openssh-server?

To get to a terminal screen from the main console press control + alt + f2.  From there you should be able to login a console and put gdm.conf back into place but you need to rename startx.conf to something else without the .conf at the end first.  Also you will not be able to use the btcminer files in /etc/init but you should be able to call startminers.sh from a regular terminal within gnome.

-Dukejer

Thanks Dukejer, once I got to the terminal I solved all the issues. Everything works fine now. Thank you very much.

[vborets]

Hi

I try use

sudo tar xfz $AMDAPPSDKROOT/icd-registration.tgz

but in etc/opencl/vendors    amd and ati files not present

i try extract its mannualy  - but not have permisions :-(

[dukejer]

Hi

I try use

sudo tar xfz $AMDAPPSDKROOT/icd-registration.tgz

but in etc/opencl/vendors    amd and ati files not present

i try extract its mannualy  - but not have permisions :-(

Hi vborets,

Did the AMDAPPSDKROOT Variable get set?

You can check it by using echo $AMDAPPSDKROOT.

Are you out of disk space?

Try running df -v to view the space.

Also did the sudo command take your password without any errors?
Sudo gives you the permissions to extract the files into /etc if you are running a user the has sudoers access rights.  Normally the first user you create in Ubuntu has sudoers access rights.

-Dukejer

If you find this post helpful feel free to donate to address: 19oQLN2v1auiGqBDFfxSDUfu8Ax6ooT6xs

[vborets]

Did the AMDAPPSDKROOT Variable get set?

You can check it by using echo $AMDAPPSDKROOT.

Are you out of disk space?

Try running df -v to view the space.

Also did the sudo command take your password without any errors?
Sudo gives you the permissions to extract the files into /etc if you are running a user the has sudoers access rights.  Normally the first user you create in Ubuntu has sudoers access rights.

-Dukejer

Problem not in disk or path- problem with permissions. I use single user

[dukejer]

Did the AMDAPPSDKROOT Variable get set?

You can check it by using echo $AMDAPPSDKROOT.

Are you out of disk space?

Try running df -v to view the space.

Also did the sudo command take your password without any errors?
Sudo gives you the permissions to extract the files into /etc if you are running a user the has sudoers access rights.  Normally the first user you create in Ubuntu has sudoers access rights.

-Dukejer

Problem not in disk or path- problem with permissions. I use single user

Hi vborets,

I am have not tested this setup in single user mode.  I originally set this up in a gnome desktop after ubuntu was installed.  

You should be able to find the .tar file using the home path where the AMD SDK extracted.

ex.  cd /
      sudo tar xvf ~/AMD-APP-SDK-v2.4-lnx64/icd-registration.tgz

or
      sudo tar xvf $HOME/AMD-APP-SDK-v2.4-lnx64/icd-registration.tgz

-Dukejer

If you find this post helpful feel free to donate to address: 19oQLN2v1auiGqBDFfxSDUfu8Ax6ooT6xs

[bgd]

Hi Dukejer,

I got to the like after the ATI driver install. Then I get:

Code:

sudo dpkg -i *.deb
(Reading database ... 67911 files and directories currently installed.)
Preparing to replace fglrx 2:8.850-0ubuntu1 (using fglrx_8.850-0ubuntu1_amd64.deb) ...
Removing all DKMS Modules

Error! There are no instances of module: fglrx
8.850 located in the DKMS tree.
Done.

This machine is setup as a server and so I haven't installed any desktop-related stuff, including X11.

I have an AMD 965 & ATI 5870, if that makes any difference.

Thanks for your help! -bgd

[dukejer]

Hi Dukejer,

I got to the like after the ATI driver install. Then I get:

Code:

sudo dpkg -i *.deb
(Reading database ... 67911 files and directories currently installed.)
Preparing to replace fglrx 2:8.850-0ubuntu1 (using fglrx_8.850-0ubuntu1_amd64.deb) ...
Removing all DKMS Modules

Error! There are no instances of module: fglrx
8.850 located in the DKMS tree.
Done.

This machine is setup as a server and so I haven't installed any desktop-related stuff, including X11.

I have an AMD 965 & ATI 5870, if that makes any difference.

Thanks for your help! -bgd

Hi bgd,

The post I wrote is for installation from a full desktop.  Even though I shutoff all the desktop process it still needs a lot of the components like Xorg to run the ATI driver.  When I originally installed the ATI driver on my original system I started with server also but I upgraded it to a full desktop by running
sudo apt-get install ubuntu-desktop.  Also before you start you  may want to make sure there are no other fgrlx packages installed on the system.

-Dukejer

If you find this post helpful feel free to donate to address: 19oQLN2v1auiGqBDFfxSDUfu8Ax6ooT6xs

[dukejer]

Hi Everyone,

I added a way in the original post a way to monitor the miners using an Xterminal on the main Xwindows console.  Here is the code for anyone interested.

How to monitor your screens on an Xterminal on the main Xwindows console on boot up.

   sudo gedit /etc/init/btcmonitor_0.conf
        description     "Start BTC Monitor 0"
        start on runlevel [2345]
        stop on runlevel [!2345]
        kill timeout 30
        script
          LOGINUSER=ChangeToMyLoginUser
          export DISPLAY=:0
          #Wait 40 seconds to make sure X is started.
          sleep 40
          exec  /usr/bin/xterm -geometry 80x40+0+0 -e "/usr/bin/screen -d -r gpu0"
        end script

sudo gedit /etc/init/btcmonitor_1.conf
        description     "Start BTC Monitor 1"
        start on runlevel [2345]
        stop on runlevel [!2345]
        kill timeout 30
        script
          LOGINUSER=ChangeToMyLoginUser
          export DISPLAY=:0
          #Wait 45 seconds to make sure X is started.
          sleep 45
          exec  /usr/bin/xterm -geometry 80x40+500+0 -e "/usr/bin/screen -d -r gpu1"
        end script

- Dukejer

If you find this post helpful feel free to donate to address: 19oQLN2v1auiGqBDFfxSDUfu8Ax6ooT6xs