Just to follow up, "1.48 debug" seemed to improve things at first for a 6950 under Win7, but I'm still not having much luck with it.  It doesn't seem to actually get stuck, but it spends most of its time idling.  It seems to eventually recover, but then it will idle again after a few minutes.

phoenix-1.48-debug>phoenix -u http://user:pass@bitcoinpool.com:8334/ -v -q 2 -k poclbm DEVICE=0 AGGRESSION=10 BFI_INT PLATFORM=1 WORKSIZE=128
...
[24/05/2011 10:42:46] Warning: work queue empty, miner is idle
[24/05/2011 11:00:13] LP: New work pushed
[24/05/2011 11:00:13] Server gave new work; passing to WorkQueue
[24/05/2011 11:00:13] New block (WorkQueue)
[24/05/2011 11:00:16] Result 00000000f651082d... accepted
[24/05/2011 11:00:18] Result 000000002db4051c... accepted
[24/05/2011 11:00:25] Warning: work queue empty, miner is idle
[24/05/2011 11:03:00] LP: New work pushed
[24/05/2011 11:03:00] Server gave new work; passing to WorkQueue
[24/05/2011 11:03:00] New block (WorkQueue)
[24/05/2011 11:03:12] Warning: work queue empty, miner is idle
[0 Khash/sec] [360 Accepted] [6 Rejected] [RPC (+LP)]

Killing it and starting fresh always seems to do well for a while, which makes me think it's not solely a pool problem.

I'm having frequent lock up problems with 1.47 on Win7 using Bitcoinpool and running on a BIOS flashed 6950.    I don't recall having these problems with some of the very early versions.  Here's the sort of messages I see when it goes off:

phoenix -u http://user:pass@bitcoinpool.com:8334/ -v -q 2 -k poclbm DEVICE=0 AGGRESSION=10 BFI_INT PLATFORM=1 WORKSIZE=128

[20/05/2011 02:00:45] New block (WorkQueue)
[20/05/2011 02:00:57] Warning: work queue empty, miner is idle
[20/05/2011 02:12:10] LP: New work pushed
[20/05/2011 02:12:10] Server gave new work; passing to WorkQueue
[20/05/2011 02:12:10] New block (WorkQueue)
....
[20/05/2011 10:17:40] Warning: work queue empty, miner is idle
[20/05/2011 10:18:03] LP: New work pushed
[20/05/2011 10:18:03] Server gave new work; passing to WorkQueue
[20/05/2011 10:18:03] New block (WorkQueue)
[20/05/2011 10:18:15] Warning: work queue empty, miner is idle
[20/05/2011 10:20:59] LP: New work pushed
[20/05/2011 10:20:59] Server gave new work; passing to WorkQueue
[20/05/2011 10:20:59] New block (WorkQueue)
[20/05/2011 10:21:12] Warning: work queue empty, miner is idle
[0 Khash/sec] [1506 Accepted] [96 Rejected] [RPC (+LP)]

This usually happens within 8-12 hours.  Closing the terminal window and restarting always seems to work, so it doesn't seem to be a pool problem. If there is a deeper level of debugging I can turn on, or something else I can do to help diagnose, I'm happy to try.  I'll also try the most current version and see if that helps.

Thanks for your effort and generosity in producing this!

I've got exactly the same problem.  New 6950 (unmodified), brand new Windows 7 install.  OpenCL appears to be working fine, as confirmed by a test of Luxmark.  Gui hangs at "0 accepted".

I initially thought the problem was conflict with other NVidia card, but that seems to be a separate issue.  Interesting symptom is that choosing CPU instead of 'cayman' produces the same result.

But the poclbm-mod.04.20.2011rev2 command line version seems to be working fine at 278Khash/s. Download, edit "sample-start" to have correct login information, and double click.