I think the more accurate answer here is not "you need to invest $1000 every 10-12 days" or "You need to add 10-15% more every 10-12 days".  The answer is that you need to figure out how much hashing power you need right now to earn 0.2BTC per day.  As has been pointed out, at current difficulty levels, that number is 7.451TH/s.  When the next difficulty jump happens, you will need to add an equivalent percentage of hashing power as the difficulty increase.  So, if the next jump is 10%, you need to up your hashing power by that same 10% (745.1GH/s) to maintain the 0.2BTC daily earnings.

If you have no hardware currently, then you're going to have to speculate what the next difficulty jump will be so that you can place your orders now to compensate.

Current pricing has 1TH/s miners (like the Dragons) at about $1500 shipped.  These miners come with power supplies and will run about 1000 - 1100 watts.
You could go the S3 route - although you'll have to wait until 8/7 for them to ship batch 5.  Each of those is 0.64BTC.  20 units should be enough, though you'd probably want to grab some extra to compensate for the difficulty jump.  These units do not come with power supplies, so you'll have to purchase those.  The Dell server supplies and breakout boards will each power 2 units.

In any case, you're looking at a decent investment up front, and then a variable additional investment every 2016 blocks.

By the way, you're going to need enough power to drive all of that hardware and enough cooling/airflow to manage all the heat they're going to generate... got your own datacenter, or planning to build one?

Huh... weird that under clocking causes you to see "x" and "-".  Even weirder that running that same unit at 218.75 is stable.  I have 1 unit that just simply won't run stable at stock clocks, but gets a pretty respectable 425-430 at 212.5.  Another runs perfectly happily at 218.75 getting 440.  Neither will over clock at all.

Your memory is failing you, then

PCI-e connectors are 6 pin or 6+2 pin.

The S1 requires 2 6-pin PCI-e connectors.  1 per blade.
The S3 at stock speeds requires the same - 2 6-pin PCI-e connectors.  1 per blade.  If you are lucky enough to get your S3s with "good" DC/DC converters, you *might* be able to stably over clock them.  If you do over clock, Bitmain states you need 4 6-pin PCI-e connectors per unit (2 connectors per blade).

As for a PSU, if you're going to power only a single S3 and not over clock it, just look for one that is at least 500W and has 2 separate PCI-e cables.

I recommend the EVGA 1300 G2.  It has 6 PCI-e cables with a total of 8 PCI-e connectors (2 of the cables are doubles).  You can easily power 3 S3s at stock clocks with it.  If you under clock and under volt your S3 (not sure if the pencil mod trick works on these like it did the S1) and can get them to around 275 - 300W each, then that same PSU will drive 4 units.

Thanks for your help.  If I understand you correctly, the following are your recommendations:

• For p2pool, queue depth of 1.
• Certain hardware that uses their own customized fork of cgminer may have different queue depth requirements due to driver implementation
• Scan time and expiry settings are pointless on modern mining hardware

Your guidance is much appreciated.

If you don't explicitly set scan/expiry parameters, what do they default to in cgminer?

EDIT: there are recommendations all over the board (bother literally and figuratively) regarding the best and/or most valuable tweaks to make.  Since you wrote the mining software, and I presume are at least conversationally familiar with Bitmain's hardware, what values would you recommend for the S3?  Should we leave it alone at Bitmain's setting of 4096?  Also, is Bitmain really using version 3.1.2 of cgminer, or have they number versions their own way because they've customized the software?  If yes, do you know the version of cgminer on which they based the one in the S3?

Thanks again!

EDIT2: Rather than just a straight up recommendation, is there some kind of formula to calculate the "best" setting for queue depth with a given hash rate and/or mining hardware?

EDIT3: Sorry!  I'm full of questions this morning... would setting the queue depth values be dependent on not only hash rate, but also the pool itself?  For example, would a miner on p2pool wish to have a value of X with hardware Y, but a miner on BTCGuild would want a value of Z for that same hardware Y.

That's an interesting point, and I probably should have considered the ramifications of setting it so low before just believing what I read as a recommendation.  I'm actually tearing down my S3s today, cleaning them completely of the messy thermal compound Bitmain used and putting AS5 on the chips.  I'll restart them with the updated scan time value of 10.

Thanks for the help.

I don't think Spondoolies-Tech has ever officially sanctioned this GB.  In fact, the only time they've actually replied in this post was to dispute claims that they were selling the SP30s at $2500 for bulk purchases.

How many units did you order?  This message means that you have to supply importer paperwork - which is nothing more than filling out a PDF that UPS will provide you - and emailing it back to them.  I had this happen when my SP10 was shipped to me.  I called Fedex about it, and the girl I spoke with said it was because the value was over $2500.

So, just call UPS, fill out the paperwork and your miners will be released and sent to you.

S3s on p2pool... flashed the new firmware this morning and edited the cgminer script to set:
First S3 hashing happily away at 440:


Second S3 hashing away at 427 (this one absolutely refuses to even hash at stock clocks without a smattering of "x" and "-":


Check out those hardware error rates... it's practically non-existent.

So it's been 12 hours give or take since I flashed the new firmware and changed the cgminer settings to queue 0 expiry 1 scan-time 1.  No real noticeable difference.

First S3 (hashes ~440):


Second S3 (hashes ~425):

You listed your problem in the first post:

IP of miners is 192.168.1.*
IP of your network is 192.168.0.*

Change your network to 192.168.1.*

Then, connect your S3 1 at a time and edit the WAN to use DHCP if you want your router to automatically assign the miners addresses... or setup the static IP addresses - ensuring each miner gets a different address (like 192.168.1.101 and 192.168.1.102).

I think it is more appropriate to state, "the amount of BTC you can expect to make is..."

It's a subtle distinction, but an important one.

Day?  No.  Since the network hash rate and difficulty do not remain static, you can't get the prediction to the day.  I can, however, tell you that the reward drops to 12.5BTC at block 420,000 and we (as of writing this post) are on 312450.

I mine on p2pool.  This morning I flashed the new firmware and set the options for --queue 0 --expiry 1 --scan-time 1.  I'll report back later with any changes (good or bad) that I notice.

Prior to making these changes, I saw ~440GH/s on one S3 with exceptionally low HW errors, WU about 6200 on stock clocks.  The other S3 refuses to even hash properly at stock clocks, so I run it at 212.5.  With that I get ~430GH/s, very low HW errors, WU about 6100.

Even funnier is the fact that the controller board of the S1/S3 is nothing more than a router itself...

You've got a couple of options:

1) Put a wireless router in the location where your miners are going to be.  Setup the router as a wireless repeater.  Plug your miners directly into the router.
2) Purchase some power line network adapter kits.  These use your in-home electrical wiring and provide you with network access anywhere there's an outlet.

I'm not sure what you mean by "control 2 miners at once".  However, if you do have multiple miners attached to your network, you either have to assign them static IP addresses (a different IP address for each miner) or set them up to use DHCP and let your router assign the addresses.

I've read a lot of misinformation on these boards and these statements are about as incorrect as you can get.  Server-grade PSUs are absolutely designed for intensive application.  They are absolutely built for 24/7 operation with no downtime.  They are purpose-built to be easily swappable in the case that they do happen to go down.

Tell the guys that run the Google data centers that their servers don't really run at high capacity.  Or the guys at Amazon.  They'll disagree pretty emphatically.

Unless you're playing your video games 24/7, you've never punished your gaming rig's PSU more than a mining rig would.  Even if you HAD played 24/7, you wouldn't have because the rendering of the game scenes and objects does not always require the same level of power supplied by the PSU.  I can guarantee you that if you throw a GPU mining rig together with the same components as you do a gaming rig, that mining rig will crash and burn first.

Overclocked an S1 is going to need about 400W.  I've been well served by Corsair and EVGA.  Personally, I'd go with the EVGA 1300 G2.  It can handle 3 overclocked S1s.

Sitting in the parking lot looking at used lottery tickets... too funny

Oh... the BTC Garden 1.2TH/s.  Yup... you're pulling about 1300 watts or more.  You're safe on either a 15A or 20A circuit.  However, you WILL NOT be able to plug that second rig in the same room.  Your 20A circuit can only handle 1920 watts total continuous load.  Also, be aware of what other equipment is in that same room and running on that same breaker.  You've got some room leftover to run some lights, laptop or low-power PC, TV etc... but definitely not another one of those mining rigs.

If you've got room in the panel, and that panel is driven by 2 120v mains (certainly should be), then your electrician should be able to throw a 240/30 in there.  Then, he can run the proper wiring and install a new "receptacle" (the nema I stated previously) for you in one of the rooms in addition to the 3 standard ones that are currently there.  Now you'll have enough power to drive 4 of those BTC Garden rigs... just need to be able to vent out all the heat they're gonna generate

There are a ton of things that can have an effect on earnings, and you've named some of them.  First, let's take the BTC -> fiat conversion out of the equation.  Given a hash rate of 3TH/s, you can easily calculate your expected daily earnings on any difficulty level.  Every online calculator out there uses the same formula to determine this value.

Assuming you started mining on 6/24...

From 6/24 - 6/28 you were mining at a difficulty of 13,462,580,115, so you would have expected to earn 0.11207BTC per day
From 6/29 - 7/11 you were mining at a difficulty of 16,818,461,371, so you would have expected to earn 0.08971BTC per day
From 7/12 - 7/24 you are mining at a difficulty of 17,336,316,979, so your current expected earnings are 0.08703BTC per day

Those numbers represent expected earnings based on your hash rate and difficulty.  How your actual earnings compare to those expected values depends on the factors such as where you're mining.  For example, if you're mining on p2pool and you're finding more than average shares, and the pool is finding more than average blocks, your actual earnings would be higher than expected.