Quote from: ancow on January 17, 2012, 06:09:34 PM

Edit: I just restarted cgminer, so we'll see what the efficiency is in a day or so, but less that 50% is quite normal here.

OK, approximately a day later. I doubled the scantime parameter and the efficiency went up to 16%. before that it stayed at ~10%.

Quote from: Hotdog453 on January 18, 2012, 02:20:04 PM

I have never seen efficiency levels that low on CGMiner. I can't fathom getting 12%.

What are you people doing on your PCs at the same time as CGMiner? I'm assuming they're not all dedicated PCs, but even on my most "inefficient" machines, I'm looking at 80+%, when they're actively being used.

The problem is that cgminer tries to keep the card active, so there is a queue of workitems. On each LP, both active and queued work is discarded. Since any merged mining pool (and apparently especially PoolserverJ with MM) will send an abnormally large amount of LPs, lots of work gets discarded and requested again right away.
There's also the fact that the default scantime is 60s, so every 60s any current work is discarded, if the GPU is too slow.

You can tweak cgminer to be more efficient even with MM and slow GPUs, but that will result in a dramatic increase in stales, so I'm not really eager to do that... Wink