Hello Folks,

I'm running cgminer 3.0.1 on Ubuntu 12.04 with AMD APP SDK 2.4.  I'm using whatever proprietary driver installs by default with either a 5450 or 6850 video card.

I have in the recent past been able to successfully use the 5450 and the above software with Slush's pool.  Although rather unimpressive at a doddering 10MHash/S, this setup does indeed work.  However, when i try this with the 6850 I have problems.

The basic problem is that although cgminer says the 6850 is humping along at about 190 MHash/S, neither Slush's pool or 50btc report any shares returned.  If I turn on debug mode while cgminer is executing, every few minutes I will see the following sequence of messages: "GPU 0 found something?" then "OCL NONCE nnnnnnnnnnnnn found in slot 0" then "GPU0 invalid nonce" and finally "Discarded cloned or rolled work."  I've been attempting to study the source code myself and have found where these messages are coming from, but do not presently understand the code well enough to solve my problem.

Although I saw the "invalid nonce" message long ago, I was hoping that given the bazillions of hashes that are being calculated, surely a handful of hardware errors would be of no concern.  However, I've since amended my guess such that this error sequence occurs when I should have a share ready to report, but something about it flunks a test, due to a hardware error.  Hence the entire share is no good and thus nothing is ever reported back to the pools I'm using.

Any insight about this would be greatly appreciated.