50 BTC if you write a complete guide on GPU mining on Ubuntu using ATI

[Mahkul]

1. The final instruction assumes you already have 'bitcoind -server' running. I hadn't realized that the RPC was not optional (and had been assuming the RPC stuff was there for people who were parts of a pool) until someone mentioned the '-server' option and I saw there was no such option in the poclbm code.

This is more miner related thing, you should have checked m0mchil's miner thread for general instructions. The tutorial was written specifically for me and I did not ask for explaining that. Its only purpose was to get me going with my ATI cards.

2. the 'export' library path stuff seems incomplete. I kept getting a libOpenCL.so.1 error with poclbm (which I knew I had installed as part of the tutorial's directions), and eventually slocate told me where it was and I did another export command which went like `export LD_LIBRARY_PATH=/usr/lib/nvidia-current/:/opt/ati-stream-sdk-v2.2-lnx32/lib/x86/:$LD_LIBRARY_PATH`

The path quoted by Raulo works perfectly for me (unless obviously you are using the 32-bit version).

3. hardwiring the paths is a little annoying because I had to rewrite all the filepaths (I'm on i386). Maybe this could be turned into a genuine pastable script

Good idea.

4. the tutorial advised me to the 2.1 SDK, but this seemed to lead to odd problems which went away with 2.2. On the other hand, I hadn't rebooted so I'm not entirely sure what was going on there.

Apparently the 2.1 SDK is the most efficient one (for mining purposes) on Linux.

[1714034917]

1714034917

[1714034917]

1714034917

[limpbrains]

I've installed everything following this guide and using Radeon HD 5870.
And started nonstop mining for about 10 hours.
Also I've tried change GPU and memory speed, but returned them to default - 850 Mhz.
After it I've turned off mining and start Ioquake3 just for testing.
After enjoing quake for about 10 minutes miner won't start. It says that I have only 1 device avaible for minig - my CPU.
So I've rebooted my PC.

And now I can't get it working. X server freezes on booting with one-color screen.
I've reinstalled Ubuntu - same rusult  
Really can't figure out what's wrong.

Nothing helpfull in Xorg logs

**UPDATE

Looks like my RAdeon don't work propertly. It succesfully started with driver after 10 minutes break.
Sorry for spamming.

[Mahkul]

I've installed everything following this guide and using Radeon HD 5870.
And started nonstop mining for about 10 hours.
Also I've tried change GPU and memory speed, but returned them to default - 850 Mhz.
After it I've turned off mining and start Ioquake3 just for testing.
After enjoing quake for about 10 minutes miner won't start. It says that I have only 1 device avaible for minig - my CPU.
So I've rebooted my PC.

And now I can't get it working. X server freezes on booting with one-color screen.
I've reinstalled Ubuntu - same rusult 
Really can't figure out what's wrong.

Nothing helpfull in Xorg logs

You re-installed complete system without problems (did it get to the "Reboot" button screen)?

[aistto]

I can't get working CL

Code:

$ ./CLInfo
terminate called after throwing an instance of 'cl::Error'
  what():  clGetPlatformIDs

$ ldd ./CLInfo 
	linux-gate.so.1 =>  (0x00144000)
	libpthread.so.0 => /lib/libpthread.so.0 (0x004d7000)
	libdl.so.2 => /lib/libdl.so.2 (0x00eaa000)
	libOpenCL.so => /home/serj/stream/ati-stream-sdk-v2.1-lnx32/lib/x86/libOpenCL.so (0x006bb000)
	libstdc++.so.6 => /usr/lib/libstdc++.so.6 (0x0034b000)
	libm.so.6 => /lib/libm.so.6 (0x00e04000)
	libgcc_s.so.1 => /lib/libgcc_s.so.1 (0x00585000)
	libc.so.6 => /lib/libc.so.6 (0x00145000)
	/lib/ld-linux.so.2 (0x00f3c000)
	librt.so.1 => /lib/librt.so.1 (0x00ba9000

I did everything by Raulo
What's wrong?

catalyst 10.10, ati stream 2.1, ubuntu 10.10, hd4830

edit
I found mistake:)
I did

Code:

tar zxfv icd-registration.tgz 

to sdk folder:)

[M4v3R]

Code:

$ ./CLInfo
terminate called after throwing an instance of 'cl::Error'
  what():  clGetPlatformIDs
Aborted

I'm struggling to get pass this error. Help, anyone?

- I have 64-bit system
- ATI Radeon 5850
- LD_LIBRARY_PATH is set
- DISPLAY is set

[Raulo]

I'm struggling to get pass this error. Help, anyone?
- I have 64-bit system
- ATI Radeon 5850
- LD_LIBRARY_PATH is set
- DISPLAY is set

Double check everything up to this point. It must work.

Is fglrxinfo giving you an error? Did you unpack icd-registration.tgz in the correct place? What is the result of

Code:

ldd ./CLInfo 

?
Check previous post for correct result.

[Mahkul]

Code:

$ ./CLInfo
terminate called after throwing an instance of 'cl::Error'
  what():  clGetPlatformIDs
Aborted

I'm struggling to get pass this error. Help, anyone?

- I have 64-bit system
- ATI Radeon 5850
- LD_LIBRARY_PATH is set
- DISPLAY is set

I was getting that after I made a typo in the Library path. Triple check it.

[M4v3R]

ldd command gives me expected output (libOpenCL.so is in the right place).
I've extracted icd-registration to /, so it created /etc/OpenCL/vendors dir and placed two files there
fglrxinfo gives me some info about the card (no error):

Code:

display: :0.0  screen: 0
OpenGL vendor string: ATI Technologies Inc.
OpenGL renderer string: ATI Radeon HD 5800 Series
OpenGL version string: 4.1.10524 Compatibility Profile Context

LD_LIBRARY_PATH looks correct, because:

Code:

$ ls $LD_LIBRARY_PATH

Gives me a list of six .so / .bc files from SDK lib directory.

[Mahkul]

ldd command gives me expected output (libOpenCL.so is in the right place).
I've extracted icd-registration to /, so it created /etc/OpenCL/vendors dir and placed two files there
fglrxinfo gives me some info about the card (no error):

Code:

display: :0.0  screen: 0
OpenGL vendor string: ATI Technologies Inc.
OpenGL renderer string: ATI Radeon HD 5800 Series
OpenGL version string: 4.1.10524 Compatibility Profile Context

LD_LIBRARY_PATH looks correct, because:

Code:

$ ls $LD_LIBRARY_PATH

Gives me a list of six .so / .bc files from SDK lib directory.

Looks to me like you omitted the ":" after exporting the path (marked in red):

export LD_LIBRARY_PATH=/opt/ati-stream-sdk-v2.1-lnx64/lib/x86_64/:$LD_LIBRARY_PATH

[M4v3R]

Yes, I omitted it to be fully certain that the path is good.
I did some fiddling here and there and I got to the point where CLInfo showed my card correctly. But I have another problem - poclbm.py doesn't see it. It displays:

• Cypress

In card list. When I pick it, it get's 0 khash/s. Anyone encountered something like this?

[Mahkul]

Yes, I omitted it to be fully certain that the path is good.
I did some fiddling here and there and I got to the point where CLInfo showed my card correctly. But I have another problem - poclbm.py doesn't see it. It displays:

• Cypress

In card list. When I pick it, it get's 0 khash/s. Anyone encountered something like this?

Cypress is the name of the GPU. Did you add -d 0 to poclbm.py?

[M4v3R]

Yes, that's when I got 0 khash/s. I also tried to run other samples from ATI SDK, all are hanging or not working.  

Edit: I switched to SDK v2.2 and I got ~700 khash/s, which is worse that my CPU. What's the matter

[Mahkul]

Yes, that's when I got 0 khash/s. I also tried to run other samples from ATI SDK, all are hanging or not working.  

Edit: I switched to SDK v2.2 and I got ~700 khash/s, which is worse that my CPU. What's the matter

Seems like using your CPU. Were you using SDK v.2.1 before?

Does ./poclbm give you two devices?

[M4v3R]

No, it lists only one. It doesn't list my CPU.

[M4v3R]

I was using v2.1 and I can't get it to work. On v2.2 I finally managed to get it working. It's doing ~250MHash, which is great . Thanks for all your help!

[Mahkul]

I was using v2.1 and I can't get it to work. On v2.2 I finally managed to get it working. It's doing ~250MHash, which is great . Thanks for all your help!

What was wrong that you were only getting 700Mhash/s?

Happy mining!

[burtyb]

I've noticed when using poclbm sometimes the GPU/CPU numbers change on subsequent runs which makes scripting interesting (to automate a miner for each GPU) - maybe that is the cause here too?

[marcus_of_augustus]

There is an omission in this tutorial that will send linux noobs astray on the driver install ...

at this point here

 

Open terminal and write the following command

$ sudo add-apt-repository ppa:ubuntu-x-swat/x-updates
$ sudo apt-get install fglrx
$ sudo apt-get install fglrx-amdcccle
$ sudo apt-get install fglrx-modaliases

This should install Proprietary ATI Catalyst drivers.

Assuming a starting position as new clean ubuntu install, between setting your repository to ppa:ubuntu-x-swat and getting the drivers you'll need to issue a command to update the repository list in the machine else you will just go ahead and install the current fglrx drivers referred to in the default repository. (10.10) which is exactly what Mahkul got.

So it should read
$ sudo add-apt-repository ppa:ubuntu-x-swat/x-updates
$ sudo apt-get update
### or alternately here you can open Synaptic and hit the "Reload" button, check ppa:ubuntu-x-swat is ticked as a repository in Settings->Repositories->Other Software ... also beware an "apt-get update will update your whole system which may take a while depending ....

and now it will get the drivers from ubuntu-xswat (if that is what you really want, I didn't use them)
$ sudo apt-get install fglrx
$ sudo apt-get install fglrx-amdcccle
$ sudo apt-get install fglrx-modaliases

Also here at the repository it recommends you completely uninstall all traces of any installed fglrx before trying to put another on the system ... and you can install a tool to do that (sudo ppa-purge ppa:ubuntu-x-swat/x-updates), so I guess they really mean it.
https://launchpad.net/~ubuntu-x-swat/+archive/x-updates?field.series_filter=maverick
http://wiki.cchtml.com/index.php/Ubuntu_Maverick_Installation_Guide#Removing_Catalyst.2Ffglrx

[mintymark]

I have followed a lot of this and have

 1 off NVIDIA GTX 470
 2 off ATI HD5850

After messing with the NVIDIA Cuda libraries (which I did in fact get working) , I gave up on this and decided to concentrate on getting the ATI cards working.

So far both the ATI cards work and give the expected hashrate, but if I run 2 threads then the combined hashrate is lower than with one card.

If both threads are started with the same -f parameter then its pretty much that the 2nd one to start has a hashrate similar to the cpu and the other one falls to about half what it was.

Otherwise it seems that the one with the lower -f parameter "wins" and the gets about half the expected hashrate, and the other does very much much worse than that.

podclm.py with no parameters gives 3 devices, the cpu, and the 2 ati cards.

Other data:
Ubuntu 10.04.
poclm version: poclbm/201103.beta1
ATI Stream SDK 2.3

Can anyone suggest what I might do to fix this?

[marcus_of_augustus]

Crossfire off?