Ubuntu Natty Narwhal 11.04 Mining Guide / HOWTO

[ryepdx]

Also, why are you using the 32 bit version of Ubuntu?  I'm not saying there aren't valid reasons to be using it, but I'm just curious if you have a 64 bit CPU, there is absolutely no reason to use the 32 bit version for mining.

If you have less than 4GB of RAM it's my understanding that a 64-bit OS will only serve to add overhead. I may be wrong about this as I haven't taken time to do much research. It's just what I've heard.

[SchizophrenicX]

I keep having problems communicating with bitcoin RPC via poclbm (not sure how to use pheonix) can someone give me a hand on this? Also, what does the flag -f# does? Higher number gives higher priority? or what?

I agree that this guide has gotta be re-written. Theres too much errors on it, especially on the x86 parts.

So you are the second person to say the guide needs to be rewritten because there are "errors" but as of yet, no one has pointed out a single error.

Also, why are you using the 32 bit version of Ubuntu?  I'm not saying there aren't valid reasons to be using it, but I'm just curious if you have a 64 bit CPU, there is absolutely no reason to use the 32 bit version for mining.

Ok I'm sorry if you feel offended, there are a lot of people including myself that have never ever used linux before. And because of that I'm saying that this guide could make do with a little consideration towards these people. It's not like oh theres a simple error here I know what to do. No I don't and I know there are people who doesn't because they PM me. I linked this guide in the Beginner's Guide and I feel that there is plausible cause for a simple revamp.

As mention, I can't point it out because I don't even know what the command are suppose to do. When I entered them, I don't even know if the output actually worked right. Although after a few days of searching the web and stuff, I got poclbm to work on Ubuntu so I can tell that there are some errors, especially on the x86 parts.

Ok I'll attempt to try and point out some of them that I think contains an error that you hasn't change yet...
(Some things u might consider: I know I'm suppose to go System and switch on some 'Universe' thing so that one of the command would go through, it's not mentioned. You made x86 & x64 download both version of stream sdk (Changed), during my x86 install, it got confusing as to which I'm suppose to install [or something] and then later on some of the stuff that requires it I couldn't get the path right since I have both and not know which to use)

4. needs the 'universe' setting change thing I think...
11. & 13. are the similar. I'm not sure if it's a required step. Although in 13, 32-bit and 64-bit is doing the same thing.
16. is really confusing is it right? I'm not sure. And I don't know where everything goes so yea... -.-|||

4. screen -S update sudo apt-get install fglrx vim openssh-server g++ libboost-all-dev subversion git-core python-numpy

32-bit 11. sudo tar xvfz ati-stream-sdk-v2.1-lnx32.tgz -C /opt
or
64-bit 11. sudo tar xvfz ati-stream-sdk-v2.1-lnx64.tgz -C /opt

12. sudo tar xvfz icd-registration.tgz -C /


32-bit 13. sudo tar xvfz ati-stream-sdk-v2.1-lnx64.tgz -C /opt
or
64-bit 13. sudo tar xvfz ati-stream-sdk-v2.1-lnx64.tgz -C /opt


32-bit 16. echo export LD_LIBRARY_PATH=/opt/ati-stream-sdk-v2.1-lnx64/lib/x86_32/:$LD_LIBRARY_PATH >> .bashrc
or
64-bit 16. echo export LD_LIBRARY_PATH=/opt/ati-stream-sdk-v2.1-lnx64/lib/x86_64/:$LD_LIBRARY_PATH >> .bashrc

I don't have all x64 machines. In fact my own PC I've been using for years is a x86, I also got a few 32-bit machines that are sitting in my uncle's store that is doing simple office stuff with PCIe slots I could use. I don't really have a need to until I came across Bitcoin where I'm investing. So I turned it into a mining rig since I'll be getting another MOBO w/ x64 processor I figured I'd use that as my desktop instead. So yes you're right, there are reasons to use it.

[ryepdx]

Hey, if it's too complicated for you, just run these scripts:

http://ryepdx.com/bitcoin/prep_up.sh
http://ryepdx.com/bitcoin/up_all.sh

You might have to type
chmod +x prep_up.sh up_all.sh
first to get them to run.

They're what I've been using to set up my 11.04 boxes. It installs the 2.3 SDK, though, which I've been led to understand is inferior to both 2.1 and 2.4. Haven't been able to get 2.4 to work on my machines yet. Probably because I'm trying to overwrite the old 2.3 install. I plan on trying from a fresh install using this guide sometime later:
http://forum.bitcoin.org/index.php?topic=10608.0

[IlbiStarz]

Instead of having to be rewritten for 32 bit, it needs to be rewritten for SDK 2.4...
Or just a separate guide.

[CubedRoot]

I ran those scripts on a 10.10 32 bit install and I am having problems across multiple GPUs.  I am running three Sapphire 5850's

When I run poclbm without any arguments it will list all of my GPUs as such:

   
      0   Cypress
      1   Cypress
      2   Cypress
      3   AMD Sempron(tm) 140 Process

So, i think great, its running! I then open a new terminal window and do the following:

./poclbm.py -d 0 --host=btcguild.com --port=8332 --user=MINER_gpu0 --pass=RPCPASS -v -w 128

It will fire up and run great, and give me decent hashrates since I have not overclocked them yet For example, I am getting about 250,000 Khash/s

So, I open up  second terminal window to get a miner running on the second GPU. I run the following in it:

./poclbm.py -d 1 --host=btcguild.com --port=8332 --user=MINER_gpu1 --pass=RPCPASS -v -w 128

It connects and works like normal, but the hashrate is off... it starts off slow, about 20,000 khash/sec, but then it will get to about 125,000 khash/sec.  Its also at this point that I notice the first miner I ran earlier is also dropping its hasrate to about the same.  Eventually both terminals will get to a point where their hasrates are equal and total up to about 250,000 Khash/sec...which is about what I get on one card!

So, I am thinking that both miners are computing on the same GPU regardless of which device I specifiy in the command.

What am I doing wrong?  

[Grinder]

11. & 13. are the similar. I'm not sure if it's a required step. Although in 13, 32-bit and 64-bit is doing the same thing.

No idea why 13 is there, it's already done in 11, and yes, the 32 bit version is wrong in 13.

16. is really confusing is it right? I'm not sure. And I don't know where everything goes so yea... -.-|||

No, its supposed to be like that, although I prefer this version:
sudo sh -c 'echo "/opt/ati-stream-sdk-v2.1-lnx64/lib/x86_64/" >> /etc/ld.so.conf.d/local.conf'
sudo ldconfig

Also, the 32 and 64 bit versions could be grouped better to avoid having so many "do either this or that" lines, and some of them could be done in one line using wildcard for the 32/64 part. Regarding all the echo lines, it's possible that people aren't in the home dir when they start. In that case it will fail miserably. Use ~/.bashrc instead just in case.

[SchizophrenicX]

wait...

32-bit 16. echo export LD_LIBRARY_PATH=/opt/ati-stream-sdk-v2.1-lnx64/lib/x86_32/:$LD_LIBRARY_PATH >> .bashrc

this is correct? so does the 32-bit do,

32-bit 11. sudo tar xvfz ati-stream-sdk-v2.1-lnx32.tgz -C /opt

or

64-bit 11. sudo tar xvfz ati-stream-sdk-v2.1-lnx64.tgz -C /opt

?

I've got no experience in Linux less so in scripting... zzz. I even had problem trying to install Win7 off my ext.HDD (doesn't boot Idk why)... But I'm trying.

Is there any simple site I can learn the basics to automating the drivers (& poclbm) install, after which I need to learn how to script poclbm to run for all my GPU and OC them -.- am I heading in the right direction with this? or are there easier ways to do it?

[Grinder]

32-bit 24. ./configure.py --cl-inc-dir=/opt/ati-stream-sdk-v2.1-lnx32/include/ --cl-lib-dir=/opt/ati-stream-sdk-v2.1-lnx64/lib/x86_32
is wrong too, should be
32-bit 24. ./configure.py --cl-inc-dir=/opt/ati-stream-sdk-v2.1-lnx32/include/ --cl-lib-dir=/opt/ati-stream-sdk-v2.1-lnx32/lib/x86_32

[Grinder]

32-bit 16. echo export LD_LIBRARY_PATH=/opt/ati-stream-sdk-v2.1-lnx64/lib/x86_32/:$LD_LIBRARY_PATH >> .bashrc
this is correct? so does the 32-bit do,

No, that line should be
32-bit 16. echo export LD_LIBRARY_PATH=/opt/ati-stream-sdk-v2.1-lnx32/lib/x86_32/:$LD_LIBRARY_PATH >> .bashrc

[befuddled]

Hi all-

So I'm about to dive into this. On the AMD website they say "Stream SDK" is now "APP SDK":

http://developer.amd.com/sdks/AMDAPPSDK/downloads/Pages/default.aspx

I have 69XX cards so I know I need 2.3 or 2.4 APP SDK.

They provide a link to the 2.4 install:
http://developer.amd.com/Downloads/AMD-APP-SDK-v2.4-lnx32.tgz

which I presume I would substitute for the 2.1 link provided in the instructions.

So when I look at the installation notes for APP SDK 2.4, it says a prerequisite is the latest Catalyst drivers, which I guess is this bit:

http://support.amd.com/us/gpudownload/linux/Pages/radeon_linux.aspx
which provides a link to a .run file. Yet your instructions don't mention downloading this first, or at all, that I can see.

Are the installation notes from AMD just wrong that I need to install these catalyst drivers first?

Also, if I were to follow these instructions, and they worked, then would I have catalyst available, with it's GUI for adjusting clock speed and such?

Thanks.

[Inaba]

Ok, I've made some changes and fixed the problems noted based on the suggestions.  What's the case for putting your stream path in local.conf as opposed to the .bashrc file?

As far as v2.4 goes... if someone wants to send me a 6xxx series card to test with, I will be happy to do it.  I don't have one currently, so I haven't gone down that road.  If anyone has a line on 6990's, please let me know, I'm looking to purchase two.

[IlbiStarz]

I have a 6870 for sale...idk if you want.

[IlbiStarz]

Hey does anyone know how to solve this:

I got a 5850, by afox, cuz my dad works for that company.
However, when I try to overclock it using the aticonfig --od-setclock=800, 1100 --adapter=3 (Which is max)
Nothing happens. I use aticonfig --odgc --adapter=all to show the clock speeds and it says current clocks : 725, 1000. Current peak : 800, 1100.
So...the peak changed but the current clock isn't changing. I used AMDOverdriveCtrl, and the "high" profile is indeed 800mhz, but it's not going there. Clicking "set" doesn't do anything either...

Help!?! Do I need to flash the bios?

[mintymark]

Cubedroot, I had this problem once,and I have no idea why. But I can say what fixed it. Use the environment variable:

export GPU_USE_SYNC_OBJECTS=1

Then fireup poclb.

regards,


  MintyMark

[marcus_of_augustus]

Hey does anyone know how to solve this:

I got a 5850, by afox, cuz my dad works for that company.
However, when I try to overclock it using the aticonfig --od-setclock=800, 1100 --adapter=3 (Which is max)
Nothing happens. I use aticonfig --odgc --adapter=all to show the clock speeds and it says current clocks : 725, 1000. Current peak : 800, 1100.
So...the peak changed but the current clock isn't changing. I used AMDOverdriveCtrl, and the "high" profile is indeed 800mhz, but it's not going there. Clicking "set" doesn't do anything either...

Help!?! Do I need to flash the bios?

Ask your dad?

[IlbiStarz]

He doesn't actually know them tho...you know what I mean? He's not actually thattt tech savvy...I doubt he would know how to oc either.

[Inaba]

You won't see your card set to peak speeds until you actually have a load put on the GPU.  You need to fire up the miner and it should bump the clock speed up to the peak.

[IlbiStarz]

I did, and it shows 725 and current clock. Something is not letting me oc it...will flashing it fix it?

[SchizophrenicX]

Ok, I've made some changes and fixed the problems noted based on the suggestions.  What's the case for putting your stream path in local.conf as opposed to the .bashrc file?

As far as v2.4 goes... if someone wants to send me a 6xxx series card to test with, I will be happy to do it.  I don't have one currently, so I haven't gone down that road.  If anyone has a line on 6990's, please let me know, I'm looking to purchase two.

Great, I'll test the 32-Bit now since I my friend just passed 1 over to me, I'll let you know if there are other stuff. Anyway the erm. set to 'Universe' thing I read is about installing libboost, I'm not sure if it's relevant but after enabling that setting installing libboost went through.

[SchizophrenicX]

I got can't locate libboost from

#screen -S update sudo apt-get install fglrx vim openssh-server g++ libboost-all-dev subversion git-core python-numpy

Then I had to go to System Settings -> Update Manager -> Settings... -> Ubuntu Software -> [Check] Community-maintained Open Source software (Universe)

After which I had this

cryptsetup: WARNING: failed to detect canonical device of aufs
cryptsetup: WARNING: could not determine root device from /etc/fstab

from

screen -S update sudo apt-get install fglrx vim openssh-server g++ libboost-all-dev subversion git-core python-numpy

I continued to follow the rest of the steps with no other notable warnings or errors until.

make -j3

this is the output

src/wrapper/wrap_cl.hpp:20:19: fatal error: CL/cl.h: No such file or directory
compilation terminated.
error: command 'gcc' failed with exit status 1
make: *** [all] Error 1

Someone wanna help?