Quote from: pekatete on September 10, 2014, 04:02:29 PM

Quote from: Biodom on September 10, 2014, 03:07:08 PM

Quote from: pekatete on September 08, 2014, 04:14:19 PM

Quote from: Biodom on September 08, 2014, 04:09:32 PM

Could you report hash difference (if any) after 24hr?

I sure will (and that is why I said "still early days ...".

Well, it has been a few days. Can you report any increase in hashing on 4.6.0 in comparison with the stock version?
Thanks.

Yep I can. I got the OK, not so good and bad news I am afraid.
1. The OK news
Upgrading to the latest cgminer clearly reduces the load and that definitely slightly improves the hashing. It may not be significant, but it certainly does.
I have not been able to collate any empirical evidence to this effect (see below), but I have no doubt cgminer 4.6 hashes faster than the stock one.
2. The not so good:
a) I have NOT been able to run a full un-interupted 24hr stretch with cgminer 4.6
b) The S3 that I updated cgminer on was overclocked to 262.5 and had been running very well without any significant HW errors.
3. The BAD
a) Having left the unit to hash as usual, I logged into the GUI the next day and was greeted with HW errors counting in the 100,000 ! In the panic, I immediately clicked the system tab and selected reboot. (I forgot to take a shot of the screen).
b) Yesterday, I gave the unit close observation for the few hours I had and the errors kept cropping up over time. This had thown me off until late last night I noticed a pattern.
The unit will hash as usual after a reboot without any issues and single digit HW errors for 2hrs. Between 2 and 3 hrs, there are two scenarios that can develop.
1. HW errors creep up into double digits and past 3 hrs they'll shoot up at will! A refresh would give any additional number from single to three digits!
2. The other scenario is one of the chips is marked X - when this happens, the HW errors remain normal BUT the hashrate decreases.
In both the scenarios above, I noticed the GUI hash-rate reduced to approx the one I had before updating cgminer at 256.25, so early this morning I decided to reduce my speed to 256.25 and so far, running like a champ! I believe by early tomorow I'll have the numbers to show against my earlier tests (barring another HW error panic ....!).

One thing that is still whirling away in my mind is why the same chip always comes trumps with the X? Also this may have something to do with power, as I am running the S3 off a 550 watt server PSU. If the unit is drawing anything over 460 watts, then that might be the reason.

Finally, I noticed that after updating to cgminer 4.6 if I do a power cycle or a reboot via the GUI, the miner status is screwed up. At times it will say its been up for hours and show a puny avg GH/s (say 12 GH/s avg) whereas I have just rebooted! But when I reboot via SSH (putty) this does not happen.

That's unusual. I've been running 4.6 for ~2 days now with no issues (clocked at 250). You can see from my hash results that there's been no significant performance difference: https://www.miningrigrentals.com/rigs/491
That small dip on the 8th was me upgrading.