Now that it's been clarified you've got a $500 budget and are looking to get the best bang for the buck, you've pretty much been given the hardware choices. At present, you can get about 3 S1s and a couple PSUs, or 1 S3 with a PSU, or one of those new tubes (although you're going to be a bit over your $500). If the $500 is an absolute limit, then you're left with the S1s or the S3.
I don't own one of the tubes, though they are certainly interesting. I particularly like the "build it yourself" aspect. I might get one just to play with it
The S1 is last-generation tech... using 55nm BM1380 chips that draw about 2W per GH/s. They were, and still are, workhorses. Solid and stable, and at present you can probably grab them for about $100. With your $500 cap, and assuming you've got to purchase everything including PSU, you'll be able to pick up 3 S1s and an EVGA 1300 G2 to power them (or you could pick up a couple Dell Server PSUs with breakout boards). Overclocked, you're going to get 600GH/s for your $500.
The S3 is Bitmain's latest rig. It uses the newer BM1382 28nm chip and draws about 0.77W per GH/s. The later batches are far more predictable and stable, unlike the earlier ones. I have units from batch 1, batch 4 and batch 5. The batch 5 units are far more receptive to over clocking. Some of the batch 1 units do not even hash at the advertised 441GH/s and had to be underclocked to run stable. At present you're going to have to find a reseller because Bitmain is sold out, and will not be sending the next batch (dubbed S3+) until the 20th of September. Like the S1, you need an external PSU, so factor that into your costs. In this scenario, if you get a good unit, you'll get 504GH/s for your $500 with some cash leftover in your pocket.
Solid answer, thanks for going into detail here. Do you think this setup will turn a profit or be more of a break-even hobby setup?
And as a follow-up to that, what do you think the minimum is that someone needs to spend to really make money mining?
Thanks in advance and I appreciate all the comments everyone else has provided so far as well, great community on here and looking forward to being a more active part of it!
Your followup questions are considerably more difficult to answer than your original query. The primary reason is that there are way too many variables involved, and a whole lot of guesswork in providing values for those variables. The biggest, and also most unpredictable, of the variables is the change in difficulty for solving blocks. In the simplest terms, as the total network hashing rate increases, difficulty increases. The difficulty is recalculated every 2016 blocks, and is used as a control to attempt to keep the block generation at 10 minutes.
Of course, the entire process of mining is nothing more than brute-forcing efforts at finding the right hash to solve the block. As such, it is entirely luck based. In other words, I have just as much chance of finding that solution on my first submitted share as I do on my hundred billionth submitted share. All we can solve for is the expected time to find that block, and the formula for doing so is as follows:
Difficulty * 2^32 / hashrate = expected number of seconds to find a block
So, to figure out our expected earnings per day, all we have to do is tweak that formula a little bit:
Block reward / (Difficulty * 2^32 / hashrate / 86400) = expected earnings per day
This is the basis for every online calculator out there. Let's give a real-world example using the S3:
25 / (23844670038 * 2^32 / 441000000000 / 86400) = 0.00930124667548
This tells us that in a perfect world, that S3 would expect to earn us 0.00930125
BTC per day. As you can also see, changes in the difficulty value are directly proportional to your expected earnings. In other words, you increase the difficulty by 10%, your earnings will decrease by 10%.
If you look at the first formula I posted, you'll see exactly how the network adjusts the difficulty - it knows the hash rate and sets the time to solve the block at 10 minutes (600 seconds) and solves for difficulty. Another thing to point out here is that the network really has no concept of actual hash rate. There is no central repository to which every single miner on the planet reports their individual hash rate. All the network knows is how long it took to solve a block. Just like it uses that same formula to calculate the difficulty, it also uses it to calculate the hash rate. It knows the time stamp of the previous block. It knows the difficulty. Therefore, it solves for hash rate.
Now that I've explained how the network operates, hopefully you can see the unpredictability inherent in it. If not, then just fall back to my original statement that as hash rate rises, difficulty goes up
. By the way, to all you purists out there, yes I realize it's a poisson distribution, but I'm trying to keep this at least moderately simple. If you want to see some really nicely done explanations of things, I suggest you read here:
http://hashingit.com. Davejh has done a great job with his articles, and I highly recommend them.
Alright, so back to your questions of profitability. The answers truly depend on you. Do you think the network difficulty is going to increase by leaps and bounds, averaging 25% per jump? If yes, then purchasing hardware is a losing proposition, even if you have completely free power. If you think the difficulty will average 10% per jump, then you'll see an ROI after 85 days or so assuming free power. No free power? At $0.10 per kWh that 85 day ROI just turned into 138. By the way, all of these are based on the assumption that you're going to plug that S3 in today. The longer it takes to get it mining, the more the profitability changes. For example, let's say you don't get it for 21 days, using that same 10% jump with $0.10 per kWh power, you're never going to make a profit.
I know this has been a long reply, and I hope it helps to give you a better understanding.
