new 6990 - problems with poclbm on 2nd GPU - you get 1st pool payout if solved

[LiSrt]

I'm new to bitcoin, and have just started mining with a new system, running Linux.

The problem is, I can't seem to use more than one GPU of the 6990 card I just bought - poclbm detects them as follows:

Code:

No device specified or device not found, use -d to specify one of the following

[0]     Cayman
[1]     Cayman
[2]     Intel(R) Core(TM) i7 CPU         970  @ 3.20GHz

when I try the first (or zeroth, if you prefer) device, everything works pretty much OK:

Code:

./poclbm.py --vectors --device=0 --verbose :@:8332
14/08/2011 09:34:31, Setting server ( @ :8332)    
/home/lisrt/lib/poclbm/BitcoinMiner.py:97: DeprecationWarning: 'enqueue_read_buffer' has been deprecated in version 2011.1. Please use enqueue_copy() instead.
  cl.enqueue_read_buffer(queue, output_buffer, output)
/home/lisrt/lib/poclbm/BitcoinMiner.py:143: DeprecationWarning: 'enqueue_write_buffer' has been deprecated in version 2011.1. Please use enqueue_copy() instead.
  cl.enqueue_write_buffer(queue, output_buffer, output)
:8332 14/08/2011 09:34:48, 9ab8eade, accepted
:8332 14/08/2011 09:35:30, b5dae8e1, accepted
:8332 14/08/2011 09:35:31, [378.608 MH/s (~0 MH/s)] [Rej: 0/2 (0.00%)]
:8332 14/08/2011 09:35:32, c63750c5, accepted
:8332 14/08/2011 09:35:46, c7b1566a, accepted
:8332 14/08/2011 09:35:51, 7c3c8bf9, accepted
:8332 14/08/2011 09:36:04, e02f1752, accepted
:8332 14/08/2011 09:36:09, d8e0e175, accepted
:8332 14/08/2011 09:36:13, 826fdfe5, accepted
:8332 14/08/2011 09:36:16, 110df020, accepted
:8332 14/08/2011 09:36:19, warning: job finished, miner is idle
:8332 14/08/2011 09:36:19, f436cb20, accepted
:8332 14/08/2011 09:36:31, [350.025 MH/s (~286 MH/s)] [Rej: 0/10 (0.00%)]

but when I try [1], it goes wrong:

Code:

./poclbm.py --vectors --device=1 --verbose :@:8332
14/08/2011 09:51:49, Setting server ( @ :8332)     
/home/lisrt/lib/poclbm/BitcoinMiner.py:97: DeprecationWarning: 'enqueue_read_buffer' has been deprecated in version 2011.1. Please use enqueue_copy() instead.
  cl.enqueue_read_buffer(queue, output_buffer, output)
/home/lisrt/lib/poclbm/BitcoinMiner.py:143: DeprecationWarning: 'enqueue_write_buffer' has been deprecated in version 2011.1. Please use enqueue_copy() instead.
  cl.enqueue_write_buffer(queue, output_buffer, output)
:8332 14/08/2011 09:52:11, 13b334a2, accepted
:8332 14/08/2011 09:52:12, Verification failed, check hardware!
:8332 14/08/2011 09:52:13, Verification failed, check hardware!
:8332 14/08/2011 09:52:13, Verification failed, check hardware!
:8332 14/08/2011 09:52:13, Verification failed, check hardware!
:8332 14/08/2011 09:52:13, Verification failed, check hardware!
:8332 14/08/2011 09:52:13, Verification failed, check hardware!
:8332 14/08/2011 09:52:13, Verification failed, check hardware!

I thought it might be a crossfire thing, so I tried disabling it - but every time I restart X, crossfire is re-enabled.  
(managed to disable it once, not quite sure how, but then poclbm couldn't see more than one GPU)


I know people can get much more with this card than the ~380MH/s I'm getting, so I'm pretty sure I should be able to use the other GPU, I just don't know where to go from here.

I'll send my first pool payout from mining on the 2nd GPU to anyone who gives the info necessary to get this all running OK.  

Any ideas?

edit - title change

[1714138684]

1714138684

[1714138684]

1714138684

[1714138684]

1714138684

[1714138684]

1714138684

[1714138684]

1714138684

[1714138684]

1714138684

[broken]

Are you using the latest ATI drivers?
I think you need at least AMD APP SDK 2.4, which supports Dual GPU Cards.
I am using Ati driver 11-5  and AMD APP SDK 2.4 with my ASUS 6990

[ohforf]

Could it be a Power supply Problem ?
This High End Card needs a damm good Power supply to run stable at maximum Speed.
A 6990 consumes up to 375 Watt... with Overclocking, its 450 Watt or more.

[Xephan]

:8332 14/08/2011 09:52:11, 13b334a2, accepted
:8332 14/08/2011 09:52:12, Verification failed, check hardware!

This seems to imply that your GPU is producing erroneous results. Since the first set went through, it might be a stability issue. One way to check is to underclock your GPU as low as it can go, set your fan speed to high, let the temperature drop low and then start poclbm only on the problem GPU. If you can get it to work like this, it pretty much points to a GPU stability issue.

[LiSrt]

Thanks to all the responses! - It's appreciated.

Are you using the latest ATI drivers?
I think you need at least AMD APP SDK 2.4, which supports Dual GPU Cards.
I am using Ati driver 11-5  and AMD APP SDK 2.4 with my ASUS 6990

I've got Catalyst 11.7 and the SDK is 2.5 - I may try downgrading to 11.4/2.4 though if that works for you.

Could it be a Power supply Problem ?
This High End Card needs a damm good Power supply to run stable at maximum Speed.
A 6990 consumes up to 375 Watt... with Overclocking, its 450 Watt or more.

I don't think this will be a problem - I've got 1200W available (one of these: http://www.silverstonetek.com/product.php?pid=284) - there's another 6990 coming in the post when Amazon have them back in stock so I've tried to make sure there's plenty of power available. I don't plan on overclocking (at least not until I've got a watercooling set-up), the only thing I've done so far is underclock the memory on both GPUs.

:8332 14/08/2011 09:52:11, 13b334a2, accepted
:8332 14/08/2011 09:52:12, Verification failed, check hardware!

This seems to imply that your GPU is producing erroneous results. Since the first set went through, it might be a stability issue. One way to check is to underclock your GPU as low as it can go, set your fan speed to high, let the temperature drop low and then start poclbm only on the problem GPU. If you can get it to work like this, it pretty much points to a GPU stability issue.

Hmm, will try that now...

[LiSrt]

OK, underclocking device 1 to the minimum possible (500) didn't make a difference, and downgrading catalyst and the SDK didn't work either (and knocked about 30MH/s off my mining), but, it seems someone else is having the same problem:
https://bitcointalk.org/index.php?topic=21687.0

Looks like changing poclbm to phoenix might work?

will investigate

[broken]

...
(and knocked about 30MH/s off my mining)
...
Looks like changing poclbm to phoenix might work?

I tried to updated my drivers to maybe get some more MHashs so i updated to:
Ati catalyst 11.7
AMD APP sdk 2.5 RC.2
pyopencl 1.2 (from 0.92)

But than i couldn't start my 2nd GPU on the 6990, or with different miner settings I could get it to start but it stayed at 0MH/s.
I tried different combinations (compiled pyopencl against app sdk 2.4 and 2.5).
When I went back to at catalyst driver 11.5 it worked again.

I'm using phoniex miner r118
./phoenix.py -u http://user:password@url:port/ -k poclbm VECTORS BFI_INT WORKSIZE=128 AGGRESSION=13 DEVICE=1

took me 1hour to get i back running ... never touch a running system

[jh1523]

My suggestion is to temporarily try BAMT https://bitcointalk.org/index.php?topic=28967.0 a USB-based Linux system dedicated for mining. It's easy to configure via one single config file, and has all the latest miners, kernels and SDKs needed. If both your GPUs work in BAMT, then it points to a problem with your Linux setup. If in BAMT the secong GPU doesn't work either then it must be an issue with the card itself.

Useful? flip to 1BFLDATk4iE6R2hYTfqNjxBZYctU9eshCM