To understand why your plan may or may not work requires you to understand the role of difficulty in your expected ROI.  Difficulty is a means by which the Bitcoin network adjusts itself to try and maintain a 10 minute block generation time.  As more and more hashing power is added to the network, blocks are solved faster and faster (typically).  Every 2016 blocks, the network looks at the average generation time and adjusts the difficulty to bring it back in line.

In a nutshell, as difficulty goes up, and your hashing power remains constant, you earn less BTC.  Here's a quick breakdown.  Your 5 S3s will hash at about 2.2TH/s.  Let's see how that would have manifested over the past few difficulties in expected earnings per day:

On 6/18, the difficulty was 13,462,580,115.  2.2TH/s expected to earn 0.08218BTC a day
On 6/29, the difficulty was 16,818,461,371.  2.2TH/s expected to earn 0.06579BTC a day
On 7/12, the difficulty was 17,336,316,979.  2.2TH/s expected to earn 0.06382BTC a day
On 7/25, the difficulty was 18,736,441,558.  2.2TH/s expected to earn 0.05905BTC a day

As you can plainly see, as the difficulty goes up, your expected earnings goes down, and how much the difficulty goes up directly corresponds to how much your expected earnings go down.  Just for fun, let's see what that same 2.2TH/s would have expected to earn in December of last year:

On 12/21/2013, the difficulty was 1,180,923,195.  2.2TH/s expected to earn 0.9369BTC a day.

So, in 7 months, that same hashing power went from earning you close to 1BTC a day to now making about 0.06BTC a day, because the difficulty went from 1.1 billion to the current 18.7 billion.  If you project that every single difficulty adjustment will be an increase of 20%, let's take a look at what that same 2.2TH/s would be earning you:

After 10 difficulty adjustments of 20% each, the difficulty would be about 116,000,000,000.  2.2TH/s expects to earn 0.009538BTC a day.  10 difficulty jumps should be right around the end of November this year.  So, at the end of a year's hashing, your gear would be earning just about 100 times less.  In fact, the decreased earnings EXACTLY corresponds to the increased difficulty.  Here's the actual formula for calculating expected earnings:


So... now you need to speculate.  What do YOU believe the difficulty will do?

We're on the same page regarding the circuit load.  Pulling the full - or close to full - load from a circuit is certainly supported for "bursts".  The microwave is a good example of this kind of load - pulling a lot of power for short periods.  What is not supported is continued and constant draw over 80%.  Your mining gear is a good example of this kind of load.

240V is always a better option simply because higher volts means less amps for the same power requirements.  Your PSUs will be happier.

The RM1000 is a good, solid choice.  I have had luck with the Corsair products, having owned a CX600M (sold it to a friend to use with his S1), and an HX1050 which powered 2 S1s for 4 months and is now powering 2 S3s.

Yeah, having about 3.6TH/s 6 months ago would have been nice (About 56BTC nice).

BTCGuild is PPLNS (I believe they had a PPS pool previously that you could mine for a fee... I'd have to let someone else clarify here)
P2Pool is also PPLNS, but works based on a share chain.  When a block is found, payouts are calculated for the previous 8640 shares.
Eligius is a variant of PPS (CPPSRB) that wizkid devised.

Why are so many people on GHash.io?  I don't know... why do so many people use Microsoft products?  Some things just defy logic

Besides the availability of such PSUs, the prices they command are unreasonable.  An EVGA 1300 G2 can be had for about $175 on Amazon.  The EVGA 1600W PSU is about $350.  I'm pretty sure I'd rather have 2 of the 1300W units for the same cost as one of the 1600W units.

Another thing to consider is the power from the wall.  If you're in the US and plugging it into your typical wall outlet, and there's nothing else at all on that circuit, you can pull a max constant load of 1440W before tripping your breaker (80% of 120V/15A), which renders the 1600W PSU a worthless investment.

All the online calculators make assumptions and use static values where dynamic ones are needed.  Another thing to consider is how you're looking at ROI.  If you purchase hardware in BTC, you need to think of ROI in terms of BTC.  If you purchase hardware in fiat (like USD), then you can think of ROI in terms of fiat.  Do not intermix the two.  Examples:

1 - You purchase Joe Miner's new rig for 3BTC.  If that miner gives you back more than 3BTC before you unplug it, you've made a positive ROI
2 - You purchase Joe Miner's new rig for $1800.  If that miner mines enough BTC that you could convert to USD to cover your investment, you've made a positive ROI.

Example of how NOT to think of things:

1 - You purchase 3BTC from an exchange for $1800.  You then use those coins to purchase Joe Miner's new rig for 3BTC.  After 6 months, the miner has mined 2BTC and you pull the plug.  In 6 months the price of BTC has gone to $1000.  You sell your 2BTC for $2000.  You DID NOT make a positive ROI.  You ended up with less BTC than you started with.

The units are hashes per second.
KH/s - kilohashes per second - 1,000 h/s
MH/s - megahashes per second - 1,000,000 h/s
GH/s - gigahashes per second - 1,000,000,000 h/s
TH/s - terrahashes per second - 1,000,000,000,000 h/s
PH/s - petahashes per second - 1,000,000,000,000,000 h/s
... and so on.

Your average PC's CPU will measure speeds in KH/s
Your average PC's GPU will measure speeds in MH/s
You average ASIC will measure speeds in GH/s - TH/s

Your last question is just another variation on your first post in this thread, "What rig will make me profit?" If we all knew what machines would guarantee profit, I'm pretty sure we'd all have bought the manufacturer out of them by now

Some people buy the miners with delusions of grandeur.  They think, "I'm gonna make a fortune in magic money!" figuring if they buy some overpriced crap on flea-bay and plug it in, they'll be flying their own private jet to their vacation island.

Mining BTC is not a get-rich-quick scheme, plain and simple.  Can mining be profitable?  Sure it can... just like any investment has the potential to be profitable.  Do your research and make your decision.

Online calculators use the following formula to figure out your expected earnings per day.  So, if you had those S3s mining right now:

The numbers aren't based on whether or not you mine in a pool.  You also need to account for the changing difficulty - which I'm sure the numbers you've provided do not.  You've based your projections on a static difficulty rate, whereas the reality is the difficulty rate is anything but static.

You need to realize that the BTC network adjusts the difficulty every 2016 blocks to try and maintain a 10 minute block generation time.  If you look at the historical difficulty charts, you'll see that it is constantly rising... sometimes ridiculous amounts... others very little.  Most online calculators will assume a 20% increase every jump by default.  Personally, I think that's not sustainable, but that's a completely different discussion

Anyway, if I use the same numbers you've given - up front hardware costs, free power, etc.. and assume you will not get your miners until next Friday at the earliest... with a difficulty increase of 20% every 2016 blocks...

By December 18 of 2015 you will have made a 0.4755BTC return.

That's the doomsday scenario where every single difficulty increase between now and then is 20%.  If you change that number to a 10% increase...

By January 9 of 2016 you'll have made a 4.436BTC return.

As you can see, there is a significant difference.  So, what do you do with all of this?  Take it with a grain of salt and decide for yourself if you want to commit your cash to mining hardware

In the long run there won't be much difference in payouts between the different pools.  Some charge fees, others don't.

You could always set your miners to load balance across a number of pools to diversify your hashing power.  Here's a good thread that shows a long-running experiment on three different pools (BTCGuild, Eligius and p2pool): https://bitcointalk.org/index.php?topic=416933.0

Somebody's got some pretty serious firepower pointed there, that's for sure.  One address has 160TH/s... the lucky guy who found the pool's only 2 blocks is there with 15TH/s.  That "high roller" would expect to find a block about every 6 days with that hash rate at current difficulty.  $17,000 a week payday.  Not too shabby.

I'm sure it did... these types of questions are typically asked by people who think they're going to get rich by leaving their 5 year old PC running in the basement and creating magic internet money.  I suppose I could have explained how I got that answer, but since the question itself shows little to no effort put in by the OP to understand how the mining process works, what role difficulty plays in the equation, or to see one of the many other threads on these forums that mention the multitude of online calculators, I figured I'd save myself the trouble and just scare him with the fact that right now he'd need nearly 2.5TH/s to mine those 2BTC a month.

Just sent you a PM... I've got 2.  One of them is my "lucky" one ... it found a block on p2pool

I can factory reset them for you, or send them to you as is (overclocked to 393.75 and running kano's cgminer binaries).

At current difficulty, 2.484TH/s

You're exactly right... the new power requirements effectively kill off any chance of running these units residentially in the US unless you've had electrical work done.  I imagine the new power requirements will also impact hosting costs as well.  All in all pretty disappointing news from the guys at SP-Tech... losing 1.5TH/s and costing 500-600W more in power.  The numbers are still pretty good in terms of efficiency, but just nowhere near what was advertised/promised/sold.  I know the SP-Tech guys have promised compensation to those who ordered, so kudos to them for at least making an effort.

I'm not sure I would call that HW error number huge... it's 0.034%.  It's certainly larger than what I experience on my S3s, which give me about 0.0017%.  The WU number is very telling, though.  You're getting 3525 and you should be somewhere between 6100 and 6200 at stock clocks.  Also, the extremely low temperature of chain one is an indication that something funky is happening there.  If you reboot it from the web console, does it come right back up and show all ASICs as "o" status on both chains?  If yes, then there's a power problem and the ASICs on that first chain aren't getting fed enough to hash at the specified clock rate.  Another question: if you power down the unit completely for about 5 minutes, then start it back up again, does it report proper hash rates, but then drops over time?  This would also point to a problem most likely in the DC/DC component.  What happens if you try to clock it lower (like 212.5 or even 200)?

Finally, I'd open a support ticket with Bitmain.

Well, Spondoolies-Tech has officially announced the production specs of the SP30:

July, August, September batches are 4.5TH/s @ 3000W from the wall
October and later are 5.5TH/s @ 3000W from the wall

Pricing will be adjusted to reflect the change in specs.  No actual pricing numbers have been presented yet, and their website has not been updated with stock/pricing, either.

Wouldn't "you lose $3.67 per day" mean "it costs you $3.67 a day to run the thing"?

The original statement is "I have a 1TH/s miner that draws 1kW of power".  We know he's using 24kW of power in a day.  If it's costing $3.67, he must be paying $0.15ish per kWh.

I am genuinely interested in how you got your number as I just can't see your interpretation of it.  Would you mind expanding for me?  Thanks!

EDIT: Never mind!  I got what you're trying to say now... your take on his statement was "At the end of the day I'm $3.67 in the hole"
So, given the current difficulty, that 1TH/s expects to earn 0.02683BTC a day, which converted to USD right now is $15.51.  So, if he's ending up at a loss of $3.67 a day, he's got to be spending $19.18 a day in power... hence your $0.80 per kWh costs.

That's kind of why most here would state it's better to use a single rail PSU.  Let me give you a couple examples:

The Cooler Master Elite V2 550W PSU is a multi-rail unit (2 12V rails) that has 2 6+2 PCI-e connectors.
The EVGA 500B 500W PSU is a single-rail unit (1 12V rail) that also has 2 6+2 PCI-e connectors.

Just from what I wrote, you would think the Cooler Master is the better PSU... higher wattage, multi-rails.  You'd be incorrect.  That Cooler Master can only draw a max of 384W and 23A from rail 1, OR 384W and 20A from rail 2.  That's right.  You've got 384W of useable power.

The EVGA 500B on the other hand gives you a full 480W and 40A off that single rail.  All of that is useable.

So... which would you choose now?  The one that will give you 480W of useable power at 40A... or the other one?

By the way, did I mention that the EVGA is also $15 cheaper on newegg.com (after rebate), and it's bronze-rated (higher efficiency)?

Err... so did you.  $0.80 per kWh translates into $19.20 a day.  To get $3.67 a day, it's $0.1529 per kWh.