Choosing a PSU when you have free power

[Witrebel]

Hey guys,

I have tried searching my question a few different ways and I haven't found much so I figured I'd post it up.   

I am looking to build a mining farm and I have access to virtually unlimited power that is for all intense and purposes free (for sure less than < 1 cent per kWh).  I understand that everyone is in the game of efficiency so you seek out the best PSU's you can find.  For me however, cost of entry is the biggest factor.  I don't care how efficient it is.  The second concern is longevity and safety. I obviously don't want to burn the place down or smoke a miner.   So in your experience what do the "cheaper" psu's bring to the table?  Can you just buy something bronze and call it a day?  Can you go no name and overspec it in terms of wattage and expect it to be safe?  Basically, if efficiency is no matter, can you recommend what PSU you would go after and where you would find them?


Furthermore, do any of you hardware guys know how clean the power needs to be?  Any hard values about max acceptable ripple would be interesting.  A part of me thinks that it may be cheaper to source parts from Digikey and build my own AC/DC converters in bulk.  Thoughts?

[1714857167]

1714857167

[1714857167]

1714857167

[1714857167]

1714857167

[1714857167]

1714857167

[alh]

What kind of AC infrastructure will you have for this "free" power?

I am guessing you don't care about noise?

Remember that an inefficient power supply will add to your cooling load, since the difference is heat.

[Witrebel]

That's a good question,  I think the first iteration of the farm would be in a more residential infrastructure with typical 110v 60hz 15amp wiring and maybe one or two 220v circuits.  After a few months of testing and proof of concept I would be pushing to move into a more dedicated warehouse type space with an industrial connection. 

Heat is again of very little concern, the space will have top of the line climate control and again I really don't mind paying for the power.  In this instance my cost of entry is much more important than the operational costs.

I am a ME/EE dual major, so while I don't have extremely indepth knowledge about power supplies and power transmission, I certainly get the basics and would be very comfortable designing and building a full wave rectified transformer based power supply.  What I don't know is how sensitive the mining equipment is to transients, start up voltages, ripple, etc. 

I have learned that DIY is often more of a pain than its worth, so ideally I would love to find out that there is some high power low efficiency turn key solution out there with acceptable reliability and safety.

[sidehack]

I'd look into retired server PSUs. There's I think four guys here on the forums selling various kits from 750W to 2900W, prices between about $30 and $200 depending on what you're getting. Almost unity power factor, very clean output DC, user-defined cabling and efficiencies from about 88 to 94 percent are common.

If you're not getting a string miner (Prisma, S5 etc) the voltage and ripple requirements aren't that stiff for most miners. But if you're looking for a long-term investment PSU you'll want good clean output because string topology is going to be much more common in the future.

[Witrebel]

I'd look into retired server PSUs. There's I think four guys here on the forums selling various kits from 750W to 2900W, prices between about $30 and $200 depending on what you're getting. Almost unity power factor, very clean output DC, user-defined cabling and efficiencies from about 88 to 94 percent are common.

If you're not getting a string miner (Prisma, S5 etc) the voltage and ripple requirements aren't that stiff for most miners. But if you're looking for a long-term investment PSU you'll want good clean output because string topology is going to be much more common in the future.

Sidehack, really happy to see you respond to this thread.  I have been following your Gekkoscience/typezero thread and its really exciting stuff.  Just when I was certain the industry was boxing out all the home/non commerical miners I stumbled across your work.  I really hope it comes to fruition!

As for my current endevour, I am still working the model, but it looks like my best chance of success will be to scoop up outdated miners that can't compete on the US grid due gh/W rates being too low and run them on cheap power.  Basically because I have low capital I am focusing on maximizing gh/$.  I started to realize that I wasn't factoring in PSU costs and this could really bite me.  But since I'm not up against the wall on optimizing my farms efficiency, I want to explore options for sourcing reliable cheap 12v.  For the time being I would be focusing on the cheapest hardware I can find, which sounds like wouldn't be string topology.

Once the first year or so is under my belt and I have an income stream, I would start investing in longer term solutions. 

Thanks for the input about server PSU's, If I could get something  at $30 per 750w I think I could really make that work.

[sidehack]

I could get you 750W PSUs for about $30 if you weren't picky about noise. Feel free to message if you want to talk details.

[alh]

I won't bore with the usual response of "There is no such thing as free electricity". That point will eventually be made to you one way or another. It will most certainly not be "virtually unlimited" as you mention.

I will however try and temper your expectations a bit. In case you haven't visited the website:

https://bitcoinwisdom.com/bitcoin/difficulty

While it's a bit dense, if you were to put together a 1 Terhash (i.e. 1000 GH) miner, it could earn about $17 in a week, for the next week. The difficulty is looking to go up before your weeks is done, and your income will probably be reduced by about 3%. This of course assumes the value of BTC doesn't change (in terms of USD). Nobody can accurately predict what will happen to the difficulty every two weeks for the next year, since it's entirely dependent on what all the miners in world do in terms of adding or removing hashrate. While the last 3-4 months have had a little increase in difficulty, I expect that situation will not last for more than another month or two. We know that a variety of manufacturers are working on next generation ASIC, and they will almost certainly manifest in terms of hashrate increases.  If you scroll down on the above website, you'll see past difficulty adjustments and when they happened. There is also a little calculator you can use to get an idea of how quickly your mining hardware will, or won't, get paid off.

And just to top it all off, sometime in the first half of 2016, your income stream will get cut in half. That's when the block reward, currently 25 BTC in a block, gets cut in half to 12.5 BTC. This happens for everybody, and it can't be stopped. It's all baked in the Bitcoin mining protocol.

While it looks like mining is "printing money", it really doesn't print it very quickly, and most attempts to speed it up are negated by difficulty increases.

If you already know all this, then you can just dismiss as cynical old guy.

[Witrebel]

I won't bore with the usual response of "There is no such thing as free electricity". That point will eventually be made to you one way or another. It will most certainly not be "virtually unlimited" as you mention.

I will however try and temper your expectations a bit. In case you haven't visited the website:

https://bitcoinwisdom.com/bitcoin/difficulty

While it's a bit dense, if you were to put together a 1 Terhash (i.e. 1000 GH) miner, it could earn about $17 in a week, for the next week. The difficulty is looking to go up before your weeks is done, and your income will probably be reduced by about 3%. This of course assumes the value of BTC doesn't change (in terms of USD). Nobody can accurately predict what will happen to the difficulty every two weeks for the next year, since it's entirely dependent on what all the miners in world do in terms of adding or removing hashrate. While the last 3-4 months have had a little increase in difficulty, I expect that situation will not last for more than another month or two. We know that a variety of manufacturers are working on next generation ASIC, and they will almost certainly manifest in terms of hashrate increases.  If you scroll down on the above website, you'll see past difficulty adjustments and when they happened. There is also a little calculator you can use to get an idea of how quickly your mining hardware will, or won't, get paid off.

And just to top it all off, sometime in the first half of 2016, your income stream will get cut in half. That's when the block reward, currently 25 BTC in a block, gets cut in half to 12.5 BTC. This happens for everybody, and it can't be stopped. It's all baked in the Bitcoin mining protocol.

While it looks like mining is "printing money", it really doesn't print it very quickly, and most attempts to speed it up are negated by difficulty increases.

If you already know all this, then you can just dismiss as cynical old guy.

I am open to and welcome any and all advice, comments, and criticism.  

I know that the power will cost me no more than 0.0048 USD per kWh, and that this number in practice will be effectively 0.0002 USD per kWh.  That is a known.  What is unknown is the infrastructure, but I am confident I can get access to at least 100amp service to start, if not 200amp.  So while it is certainly not FREE and obviously I would need industrial grade power for an industrial sized rig, I consider it low enough to be a non/small factor for startup.  

As for the difficulty and block halving, its certainly a concern.  Does 5% monthly seem reasonable for a prediction?

Furthermore I certainly expect the network to expand again with the next gen of asics, but I don't think it can expand much faster after this round.  We are pushing the limits of the fab houses.

I have been working on a little python script to evaluate the ROI on equipment.  Code posted below if you care to look, I am sure its not accurate yet, this was my first draft:

#BTC monthly return calculator
 

ghrate     =    1000                    #rate in GH of equipment
cost       =    300                       #cost of one unit
pcons      =    1000                      #power consumption in Watts
detail     =    24                      #number of months to detail
annual     =    1                       #flag to set if you want to see monthly report or annual report



mdelay     =    2                       #number of months till the unit will see service
currentdif =     47589591154             #get todays difficulty
prediction =    1.05                    #constant percentage for monthly difficulty increase (THIS IS IMPORTANT)
erate      =    0.0048                  #cost in KWH of electricty
m2g        =    15                      #of months till next block halving
brs        =    25                      #starting block reward
b2d        =    220                     #dollars per bitcoin
bpred      =    1                       #bitcoin conversion prediction rate

mreturn = 0 #container for monthly return
mprofit = 0 #container for monthly profit
oprofit = 0 #overall profit
oreturn = 0 #overall return

tdiff = currentdif
trate = b2d



oreturn = oreturn - cost #factor in cost

mpcost = (pcons/1000)*erate*24*30 #caluclate fixed monthly energy cost
print("Monthly Power Cost =",mpcost)
tpower = 0


if(annual == 0):    #if annual is zero, show monthly details
    for i in range(1,1200000000): #just a large for loop, arbitrary upper limit
        
        if(i == m2g): brs /= 2 #if we are past the block halving period, cut the reward in half

        if(i > 1): #if its been more than a month, up the difficulty
            tdiff *= prediction #multiply the difficulty by our predicition
            trate *= bpred #same for bitcoin rate
            
        if(i > mdelay): #if we are passed the delay to implement the hardware, start counting
            mreturn = (b2d * ((30 * brs * (ghrate*1000000000) * 86400) / (tdiff * 2**32))) - mpcost #prediction line
      
            
            oreturn = oreturn + mreturn
        if(i < detail): #if we are past the range of detail we want to see, start counting till we are no longer profitable
            if(mreturn > 0):
                print("Month:",i,"Return:","%.2f" % mreturn,"Overall:","%.2f" % oreturn,"Difficulty (B):","%.2f" % (tdiff/1000000000),"BR:",brs,"BRE","%.2f" % trate)
            if(mreturn < 0):
                 print("Month:",i,"Return:","%.2f" % mreturn,"Overall:","%.2f" % oreturn,"Difficulty:","%.2f" % (tdiff/1000000000),"BR:",brs,"BRE","%.2f" % trate)

                 print("Total return after",i,"months was:","%.2f" % oreturn,"Total power cost:","%.2f" % (mpcost*(i-mdelay)))
                 break
        elif(mreturn < 0):
            print("Total return after",i,"months was:","%.2f" % oreturn,"Total power cost:","%.2f" % (mpcost*(i-mdelay)))
            break
else: #show it as an annual breakdown
    annual = []
    atemp = 0
    breakeven = 0
    for i in range(1,1200000000):
        if(i == m2g): brs /= 2 #if we are past the block halving period, cut the reward in half
        if(i > 1):
            tdiff *= prediction #multiply the difficulty by our predicition
            trate *= bpred #same for bitcoin rate
                
            if(i > mdelay): #if we are passed the delay to implement the hardware, start counting
                mreturn = (b2d * ((30 * brs * (ghrate*1000000000) * 86400) / (tdiff * 2**32))) - mpcost
                oreturn = oreturn + mreturn
                if(breakeven == 0 and oreturn > 0): breakeven = 1
                atemp = atemp + mreturn
            if((i%12) == 0):
                annual.append(atemp)
                atemp = 0
            if(mreturn < 0):
               break
    print("Break even in",breakeven,"months. Total ROI will be","%.2f" % oreturn,"in",i,"months","Total power cost:","%.2f" % (mpcost*(i-mdelay)))
    for i in range(0,len(annual)):
        
        print("Year",i+1,"return:","%.2f" % annual)
              
        

I am not a programmer so please excuse the dirty code


And sidehack, I appreciate the offer and when I am in the procurement phase I will definitely be in touch, I would like to run as much of my purchases as I can through people like you who are actually contributing to the community.

[alh]

The difficulty adjusts every 2016 blocks. If things are stable, that works out to 14 days. If the hashrate is increasing, then it will finish the 2016 blocks in less than 14. The sooner those 2016 blocks are finished, the sooner, and greater, the difficulty will increase. In a sense, difficulty rises faster (< 14 days) than it falls (> 14 days). The halving as I recall will happen after 410,000 blocks are done, and we around block 380,000? You can confirm these values with some Google research.

So you will see roughly 2 difficulty changes in a month, so 5% works out to 2-3% for a change. You can see the historical values, though you are right that the rate of increase has been slowing. 2014 was a brutal year for difficulty increases, and it probably won't be as brutal this year. So far it's been fairly benign.

A lot of this will be driven by the price of Bitcoin, and I have no idea what actually drives that. One thing I am pretty certain of is that difficulty follows price. If the price increases, difficulty follows suit. If we see $300-400 BTC price, a whole lot of hashrate will "appear" in short order. It won't be all new gear, some will be older less profitable gear that's just powered off. I think there is a chunk of currently "cold" hashrate, waiting for a better BTC price.

Looks like you are going to try and turn some of that hardware on, after it's been sold to you.

[Witrebel]

The difficulty adjusts every 2016 blocks. If things are stable, that works out to 14 days. If the hashrate is increasing, then it will finish the 2016 blocks in less than 14. The sooner those 2016 blocks are finished, the sooner, and greater, the difficulty will increase. In a sense, difficulty rises faster (< 14 days) than it falls (> 14 days). The halving as I recall will happen after 410,000 blocks are done, and we around block 380,000? You can confirm these values with some Google research.

So you will see roughly 2 difficulty changes in a month, so 5% works out to 2-3% for a change. You can see the historical values, though you are right that the rate of increase has been slowing. 2014 was a brutal year for difficulty increases, and it probably won't be as brutal this year. So far it's been fairly benign.

A lot of this will be driven by the price of Bitcoin, and I have no idea what actually drives that. One thing I am pretty certain of is that difficulty follows price. If the price increases, difficulty follows suit. If we see $300-400 BTC price, a whole lot of hashrate will "appear" in short order. It won't be all new gear, some will be older less profitable gear that's just powered off. I think there is a chunk of currently "cold" hashrate, waiting for a better BTC price.

Looks like you are going to try and turn some of that hardware on, after it's been sold to you.

Exactly, playing the arms race with new ASICS is a rich mans game.  I am not a rich man.  But as far as I can tell, there is nothing "wrong" with the older hardware other than it being too inefficient for todays prices based on power consumption.  The next generation of hardware may even strengthen my position as it might push the older units further out of the picture and thus lower their price.   You raise a fair point though about how it all ties back into difficulty.  That is certainly a feedback loop worth investigating and trying to model.

[jonnybravo0311]

I won't bore with the usual response of "There is no such thing as free electricity". That point will eventually be made to you one way or another. It will most certainly not be "virtually unlimited" as you mention.

I will however try and temper your expectations a bit. In case you haven't visited the website:

https://bitcoinwisdom.com/bitcoin/difficulty

While it's a bit dense, if you were to put together a 1 Terhash (i.e. 1000 GH) miner, it could earn about $17 in a week, for the next week. The difficulty is looking to go up before your weeks is done, and your income will probably be reduced by about 3%. This of course assumes the value of BTC doesn't change (in terms of USD). Nobody can accurately predict what will happen to the difficulty every two weeks for the next year, since it's entirely dependent on what all the miners in world do in terms of adding or removing hashrate. While the last 3-4 months have had a little increase in difficulty, I expect that situation will not last for more than another month or two. We know that a variety of manufacturers are working on next generation ASIC, and they will almost certainly manifest in terms of hashrate increases.  If you scroll down on the above website, you'll see past difficulty adjustments and when they happened. There is also a little calculator you can use to get an idea of how quickly your mining hardware will, or won't, get paid off.

And just to top it all off, sometime in the first half of 2016, your income stream will get cut in half. That's when the block reward, currently 25 BTC in a block, gets cut in half to 12.5 BTC. This happens for everybody, and it can't be stopped. It's all baked in the Bitcoin mining protocol.

While it looks like mining is "printing money", it really doesn't print it very quickly, and most attempts to speed it up are negated by difficulty increases.

If you already know all this, then you can just dismiss as cynical old guy.

I am open to and welcome any and all advice, comments, and criticism.  

I know that the power will cost me no more than 0.0048 USD per kWh, and that this number in practice will be effectively 0.0002 USD per kWh.  That is a known.  What is unknown is the infrastructure, but I am confident I can get access to at least 100amp service to start, if not 200amp.  So while it is certainly not FREE and obviously I would need industrial grade power for an industrial sized rig, I consider it low enough to be a non/small factor for startup.  

As for the difficulty and block halving, its certainly a concern.  Does 5% monthly seem reasonable for a prediction?

Furthermore I certainly expect the network to expand again with the next gen of asics, but I don't think it can expand much faster after this round.  We are pushing the limits of the fab houses.

I have been working on a little python script to evaluate the ROI on equipment.  Code posted below if you care to look, I am sure its not accurate yet, this was my first draft:

Code:

#BTC monthly return calculator
 

ghrate     =    1000                    #rate in GH of equipment
cost       =    300                       #cost of one unit
pcons      =    1000                      #power consumption in Watts
detail     =    24                      #number of months to detail 
annual     =    1                       #flag to set if you want to see monthly report or annual report



mdelay     =    2                       #number of months till the unit will see service
currentdif =  	47589591154             #get todays difficulty 
prediction =    1.05                    #constant percentage for monthly difficulty increase (THIS IS IMPORTANT)
erate      =    0.0048                  #cost in KWH of electricty
m2g        =    15                      #of months till next block halving
brs        =    25                      #starting block reward
b2d        =    220                     #dollars per bitcoin
bpred      =    1                       #bitcoin conversion prediction rate

mreturn = 0 #container for monthly return
mprofit = 0 #container for monthly profit
oprofit = 0 #overall profit
oreturn = 0 #overall return

tdiff = currentdif
trate = b2d



oreturn = oreturn - cost #factor in cost

mpcost = (pcons/1000)*erate*24*30 #caluclate fixed monthly energy cost
print("Monthly Power Cost =",mpcost)
tpower = 0 


if(annual == 0):    #if annual is zero, show monthly details
    for i in range(1,1200000000): #just a large for loop, arbitrary upper limit
        
        if(i == m2g): brs /= 2 #if we are past the block halving period, cut the reward in half

        if(i > 1): #if its been more than a month, up the difficulty 
            tdiff *= prediction #multiply the difficulty by our predicition
            trate *= bpred #same for bitcoin rate
            
        if(i > mdelay): #if we are passed the delay to implement the hardware, start counting
            mreturn = (b2d * ((30 * brs * (ghrate*1000000000) * 86400) / (tdiff * 2**32))) - mpcost #prediction line
       
            
            oreturn = oreturn + mreturn 
        if(i < detail): #if we are past the range of detail we want to see, start counting till we are no longer profitable 
            if(mreturn > 0):
                print("Month:",i,"Return:","%.2f" % mreturn,"Overall:","%.2f" % oreturn,"Difficulty (B):","%.2f" % (tdiff/1000000000),"BR:",brs,"BRE","%.2f" % trate)
            if(mreturn < 0):
                 print("Month:",i,"Return:","%.2f" % mreturn,"Overall:","%.2f" % oreturn,"Difficulty:","%.2f" % (tdiff/1000000000),"BR:",brs,"BRE","%.2f" % trate)

                 print("Total return after",i,"months was:","%.2f" % oreturn,"Total power cost:","%.2f" % (mpcost*(i-mdelay)))
                 break
        elif(mreturn < 0):
            print("Total return after",i,"months was:","%.2f" % oreturn,"Total power cost:","%.2f" % (mpcost*(i-mdelay)))
            break
else: #show it as an annual breakdown 
    annual = []
    atemp = 0
    breakeven = 0
    for i in range(1,1200000000):
        if(i == m2g): brs /= 2 #if we are past the block halving period, cut the reward in half
        if(i > 1):
            tdiff *= prediction #multiply the difficulty by our predicition
            trate *= bpred #same for bitcoin rate
                
            if(i > mdelay): #if we are passed the delay to implement the hardware, start counting
                mreturn = (b2d * ((30 * brs * (ghrate*1000000000) * 86400) / (tdiff * 2**32))) - mpcost
                oreturn = oreturn + mreturn
                if(breakeven == 0 and oreturn > 0): breakeven = 1
                atemp = atemp + mreturn
            if((i%12) == 0):
                annual.append(atemp)
                atemp = 0
            if(mreturn < 0):
               break
    print("Break even in",breakeven,"months. Total ROI will be","%.2f" % oreturn,"in",i,"months","Total power cost:","%.2f" % (mpcost*(i-mdelay)))
    for i in range(0,len(annual)):
        
        print("Year",i+1,"return:","%.2f" % annual[i])
               
        

I am not a programmer so please excuse the dirty code


And sidehack, I appreciate the offer and when I am in the procurement phase I will definitely be in touch, I would like to run as much of my purchases as I can through people like you who are actually contributing to the community.

I threw the appropriate tags around the code for you...

Also, I wouldn't use a "months until halving" variable.  We know exactly when the next halving will occur: block 420000.  With your electricity costs virtually negligible, you can pretty much put up any hardware you want.  Obviously, the more efficient it is, the more hash you can get; however, with the older generation gear selling for virtually nothing, you could pick some of it up for dirt and make some coin to use when the next gen gear comes out.

By the way... feel like hosting some gear for me?

[Witrebel]

Yeah, I get the feeling that at this price for power almost anything will turn a profit.  The question is, how long will it take to ROI, and how long will it stay profitable.  The last big obstacle to work out is shipping costs and customs issues, but I have the right resources to navigate that those issues.  To be perfectly honest the right business approach here would absolutely be to get into the hosting game. But until I have proved this is a viable mining location by actually mining with my own money and working out the supply chain/import process, I wouldn't feel right handling other peoples gear.   That said, if my first/residential phase proves successful and I get into a larger warehouse type space, I would be very open to setting up a hosting service.  But that's a LEAST a year or two out, and who knows what will happen in that time.

Also, as for the months till halving variable, I couldn't see a quite and obvious method of tracking "blocks" solved, iterating monthly seemed to be a much quicker implementation. But I suppose I could just rewrite it to iterate biweekly and assume 2016 blocks per iteration.  Thoughts?

[jimmothy]

I know that the power will cost me no more than 0.0048 USD per kWh, and that this number in practice will be effectively 0.0002 USD per kWh.  That is a known.  What is unknown is the infrastructure, but I am confident I can get access to at least 100amp service to start, if not 200amp.  So while it is certainly not FREE and obviously I would need industrial grade power for an industrial sized rig, I consider it low enough to be a non/small factor for startup.  

Are you 100% sure about that? It sounds like you might be confusing the $/kwh with the transmission fee or something.

$0.0002/kwh is impossible for any profit driven electric company. Even $0.005/kwh is hard to believe considering most industrial electricity around the world is >$0.03/kwh.

[Finksy]

The best calculator I have found to try modeling ROI is  https://bitcoinwisdom.com/bitcoin/calculator     Personally, I believe sticking to 2% per difficulty increase is still reasonable for the short term.

Using market prices for used miners (your best pricing will be on this forum typically), you can see both how quickly you can ROI, as well as how much each miner will make you in a determined timeframe. You might find that spending more upfront (and taking longer to ROI) will eventually provide you more in the long run.  Hash rate is hashrate, whether it's efficient or not is the question. My suggestion with extremely low-costing electricity like you are describing would be to try to maximize your power infrastructure with the least amount of capital as possible, and slowly replace it with newer models. One thing to take into account when considering purchasing hardware though, miners that are considered old right now will be essentially worthless by the next generation (and therefore have no value for equity), however newer miners today can often be re-sold for some portion of their purchase value in the future. The mining calculators do not take that into account, as it's hard to predict. It has been my experience that selling miners off before they are obsolete (possibly even before they ROI) can net you more in revenue than keeping them until they have depreciated completely. You need to always be evaluating your equipment, and not what it has done thus far, but what it will do moving forward (whether that is mining or selling it)

Much like investment portfolios, everybody's situation, assumptions and expectations are different, and the best thing to do is find out all the information required to make your own decision.

[alh]

I will ask a question that's been bugging me for a while:

Where are you located in the world?

I have no desire to pry into your personal life, nor am am I going to relocate to mine Bitcoins more efficiently. You electric rates are however are WAY lower than I've heard of anywhere.

[Witrebel]

@Jimothy and alh

I am currently located in the Northeast USA.  I have some close personal connections to Venezuela, and that is where I intend to set up a mining operation.  The rarity of US dollars and bitcoins in the area allows me to leverage exchange rates and lower the effective energy costs further.  For the short term I will probably do remote management on a smaller (<10 Th) farm to find out what we are up against with customs and importing this stuff, as well as how the connectivity is.  At some point I would probably move there and have an associate here in the US handle procurement and shipping of new hardware.
 
@Finsky

Really appreciate the advice.  After running the numbers I came to the same conclusion: Older gen hardware will stop turning reasonable profits inside of 2 years, at which point there will be virtually 0 resale value.   I think your idea about reselling prior to obsoletion makes alot of sense.  I also have some altruistic notions about reselling/donating older gear to the local communities to help them fight the economic disparity and lack of internationally exchangeable currency. But your point is well heard about spending a little more on the newer generations of hardware so that it stays relevant longer.  To that point, now that we are bumping up on the 16/14nm limit, do you really see the difficulty doubling any time soon?  What are you guys' thoughts on the future of the difficulty trend? 

[sidehack]

You should have a chat with chiguireitor; I believe he's from Venezuela and is running a fair bit of equipment. He could probably be a good resource for knowledge.

[alh]

While it seems to me that difficulty will take a while to double, it will continue to rise it seems. There doesn't seem to be any obvious reason for it to plateau unless/until ASIC development ceases for more efficient hardware. I guess if the price of Bitcoin falls enough, there won't be any money to fund the development of new ASICs. Of course you read about companies like 21 Inc that want to push out tons of micro-miners with a fairly efficient profile. If that actually happens, difficulty will rise again.

One thing you may have already considered. If the infrastructure in Venezuela isn't stellar, you'll want to send in only robust and reliable equipment. That includes not only the miners themselves (i.e. the ASIC gear), but the PSU's as well as any required networking switches, routers and such. It could really blow a hole in any projections you make if you lose hunks of your mining operation due to a piece of gear that isn't as obvious as the miner itself. You won't need top of the line new PSU's or network stuff, but the lowest initial price may well cost more in the end. It's the non-glamourous stuff that could really put a crimp in you plans.

Just my two cents. I hope you already speak Spanish?   

[Finksy]

I have a customer setting up a farm in Venezuela.  I believe import taxation is very heavy in that country, so I would shy away from having anything drop-shipped to your hosting location from the manufacturers, and instead opt to arrange shipping yourself. IMO server PSU's are the only way to go, there are many options to choose from on this forum. With your price of power you are not bound to look for the most efficient PSU's (as the additional cost of getting higher efficiency may not be worth it), but knowing your infrastructure limitations will help you decide if you'd benefit from squeezing more miners on-line due to lower power consumption.

One last thing to take into consideration, as has been mentioned, most bitcoin calculators do not factor in the block reward halving which will take place next year.  Personally, if I was setting up a new farm in another country, I would either work to have it operational ASAP and re-sell the hardware before the halving takes place to maximize short-term revenue (equipment will be worth considerably less when it only earns half as much as it did the week prior), or plan for the very long term factoring the reward halving in.