I would remove the -pll 850 parameter. I am getting appropriate hashrate without using the -dualminer-pll parameter, and I have noticed instability when using this parameter on previous versions of cgminer. Regarding your invalid nonce error, I get those too occasionally, though you should see less once you drop your pll parameter. Are you mining over a wifi connection? This would be the next thing I investigate if removing the non-necessary parameters doesn't help and you are using 9 DMs per p10-u2.
My rig is on a direct Ethernet connection, not Wifi. I'll take the parameter out and keep any eye on the logs.
Thanks CruzCoins!
The HW error is expected, especially, when DM works above 850Mhz.
The HW error means a returned nonce result does NOT meet the mining target requirement (if you know some mining theory) . This could happen in some cases:
1. The chip could do a wrong calculation. If it is overclocked, the chip could generate more error. That's why 850MHz will give more errors than 800Mhz.
2. Some errors could happen in the communication between DM device and cgminer. Something like uart buffer overflow.
Since higher clock frequency could introduce instability of the electronic signal, most of HW error is caused by the first reason.
HW error is different to the reject error. Reject error is caused by mining pool. If a new block had been generated and DM was still working on previous job, some nonces of previous job maybe still queued in a buffer of cgminer. If cgminer submit these old result to pool, the pool will reject them.