That's an awesome accomplishment, and I hope you can do it again here.  I am bouncing my S5 here and another small pool, it would be great for the little guys to get one.

RE: the Avalon 6 group buy, that's a great deal for a great machine, and it comes plug and play - you can't beat that.  Mine runs right @ 0.3 w/GH, and is quieter (with the Sanyo fan) and cooler than my S5.  Highly recommended.

I run my S3+ at 250 also, it stays solid at 500GH.  Runs really well with kano's updates software package, too.  I will be hosting some gear pretty soon, and when I do I'll bring the S3+ back out because you can just set it and forget it.

4.9.0 is not the version bitmain forked (https://en.wikipedia.org/wiki/Fork_%28software_development%29 - fork means they took the cgminer code and reworked it to their own ends, but along the way they messed stuff up and did things inefficiently and ignored feedback from ck and kano), I think they forked like v4.6.0 or 3.8.0 or something.  There is a whole history of bitmain's cgminer crappy forks, a little searching on the forum should shed some light (especially posts by ck and kano on the subject).

Anyway, no reason not to use the latest 4.9.2, you should actually see improved performance.  Just don't use 4.9.1, it had a bug where it wouldn't recognize the U3.

U3 thread (every U3 owner should ready this thread page 1-end): https://bitcointalk.org/index.php?topic=827356.0
Things bitmain's cgminer fork ignores: https://bitcointalk.org/index.php?topic=294499.0
cgminer 4.9.2 stuff (can read through pages following for occasional U3 related posts): https://bitcointalk.org/index.php?topic=28402.msg11597292#msg11597292

The U3 zombies no matter what software controls it, it's not a cgminer issue and definitly not a v4.9.2 issue.

Bitmains fork of cgminer is crap, it's filled with bugs and crappy code and is outdated.  I invite you to do a little research on it.

Why would you want to use outdated and broken software?

Thanks for the quick answer about 'N', how about shift information?  I can't seem to find it anywhere on the site, thanks.

I would like to request that BCMonster also be added to the pool list, http://www.bcmonster.com/index.php.  Both BC and Bravo use the same pool code, so adding one should be as easy as adding both.  Thanks!

Wow, I can't believe I have missed this thread until now.  Looks like this is replacing the gekko science pod miner then?  Way cool, I will follow this thread anxiously as I have both and S1 and S3 that I am not using, as well as an S5 that I was going to have hosted but will now hold off until these update boards are ready.  Wicked.

Looks like an interesting project if it comes to fruition, posting here to subscribe to thread.  Hopefully we'll see a working prototype.

I use two 850w gold rated OCZ psus for mine because that's what I happened to already have on hand.  Overkill like you, but better than not enough wattage as the psus will run more efficiently at lower load.

I have the Sanyo fan, too, and am using 40 for fan.

You are the exception, not the rule, and illustrate that every U3 is different.

You are using an old bitmain fork of cgminer if you're still using bmsc options, try updating to 4.9.2, it has fixes specific to the U3 in it (and uses --au3 commands instead).

Why does he have so many connections?  Seems like if you were running 300+ TH you would know how to connect to pools.

Holy cow, you keep your house like a refrigerator!  My AC is set to 78*, it would run 24/7 for almost the entire year at your temp!  Hell, my heater comes on at less than 73*...

Priority 1 should be to remove the hot air, cool air supply should be #2.

Someone hack into your machine and change the password and point workers toward their own account maybe?

Welcome to kano.is, I hope you enjoy your stay.  It surprises me that miners of substantial size are only just now finding this pool.

What is this Bitfury upgrade for the S5, do you have a link about this to share?

I may be misunderstanding your intent with this statement, but it is not a fallacy and in fact the sole and only thing you can do to increase your chance of solving a block is to add more hash power.

I do not think a 1000w psu is enough for an A6 (although I believe the A6 with downclock itself based on available voltage), mine regularly sees 1000w at the wall.  You don't want your psu % usage to be more than it's efficiency rating, i.e. a gold psu should only use 80% of it's rated power at max. continuous load, at least that's how I've always operated.

Otherwise, the "hash in waves" phenomena is exactly what mine was doing when I had its exhaust flow too restricted.

Regardless of who is paying for it (your landlord in this case), a more efficient machine would serve you better.  Get a used S3 (they're cheap!) and for the same wattage you get 2-3x the hash rate.

If I understand PPLNS correctly, which I doubt I do, you would be extending the number of payable shifts, but each shift would also be worth less, so your net in the end would be the same (?).  Is that correct?  I don't think there's a way for you to magically get paid more BTC for doing the same amount of work.

How long are you letting it run for?  Most pools start out with a diff around 1000, and it's going to take a long time for 5GH to submit a couple shares >1000 diff to get the pool to adjust the worker diff down.  Until that happen it will only be submitting shares less than the pool diff so they will be rejected.

Nice, but you should spin it 180* so that it has room to vent the exhaust air out and not right into a wall.