Clearing up some misconceptions about Bitcoin mining costs

[Raize]

I see people talking about how widespread acceptance of Bitcoin is going to increase the price of Bitcoins. It does, but it's mostly a temporal effect, especially in regards to how quickly difficulty will skyrocket to account for this increase. Over time, the price of Bitcoin will always fall back to a slight premium over the cost of mining the coin, this is because miners must profit in order to mine, but the cost of entry will be low enough otherwise no more miners will enter the system.

I know what you're thinking right now, the price of BTC has risen though from $.01 to $32, if there was widespread adoption it could go right back up to $32, right? No, the widespread adoption or demand would have to be significantly higher than the number of new miners entering the market who can make a profit even at $4. It went from $.01 to $.85 solely due to GPU mining becoming viable. It went to $2 because of difficulty increases, and it went to $32 on speculation alone. That was entirely a manufactured increase that was certain to fall over time as miners accumulated and began selling their mined coin.

If you are living in Europe or on the coasts of the US, you're now probably thinking, "But no one can make a profit at $4!" And you'd be wrong. Those of us in the US Midwest enjoy a $.08/kilowatt-hour cost, and are making nearly a $2.50/coin profit at present prices ($4.50). Additionally, those of us with extremely power efficient GPU miners can very nearly match those of you running FPGA miners even if the price keeps falling by extreme underclocking of our GPUs. In a worst-case scenario, we can buy FPGA miners ourselves as well.

I guess the point I'm making right now is, if you are getting charged more than $.08/kWh (some parts in South Dakota I know charge $.06/kWh), you are better off buying coin (or waiting and buying coin) than you are mining. You're going to ultimately lose money as those of us with the lower costs are selling our coins as we mine them. I don't want to betray a good friend's trust, but I can tell you that there is a very large operation in Kansas City that sells coin as they mine and there is no way you guys are going to run them out of business with FPGA mining in Europe, the price of Bitcoin is going to continue to fall as long as it is profitable for them to sell, and it is still very profitable for them to sell right now. It was so profitable for them to sell at $32, that they were able to reinvest into more and more. They were still doing this at even $8/coin when the rest of the forums were whining about how it was unprofitable to mine and they were enjoying a revenue/cost ratio of over 4:1.

TLDR;I would highly recommend that you not get started in Bitcoin mining if you can't make a profit at a $3.00 Bitcoin, because there are going to be players that can and are.

[mndrix]

The price of Bitcoins is set by supply and demand interactions in the market.  Bitcoin supply is fixed and predictable, so demand is the major factor in establishing Bitcoin prices.  Once price is set by the market, miners enter or leave the network based on their underlying costs.  This changes the difficulty, but not the market price.

In summary: demand leads price leads difficulty

[koin]

Over time, the price of Bitcoin will always fall back to a slight premium over the cost of mining the coin

not quite.  actually, not at all.  stated more correctly -- the amount of mining occurring will grow or shrink to a point where it is just marginably profitable to mine.

that is why when you view a difficulty versus price chart ( http://bitcoinx.com/charts/chart_large_log.png ), it is price that leads and difficulty (which reflects the amount of mining occurring) that only follows.  also this is explained here: http://virtuallyshocking.com/2011/05/25/the-relationship-between-bitcoin-price-and-difficulty/

[Gabi]

There is a maximum amount of bitcoins, 21 millions. If more people start using them, then no matter what miners do, there will be less bitcoins for everyone. So, their price will increase.

[cablepair]

My entire mining operation was paid for by mining profits starting when bitcoin was $0.85 usd, I can trace my entire operation down to an initial $150 cash investment ( a single used 5870) the rest was paid for with mining. I'm not going to go into detail but I make about double what my entire electrical bill is.

If you want to invest in Bitcoin don't mine or buy coins, invest in one of the great Bitcoin businesses or start one of your own. Cheers!

[Dargo]

those of us with extremely power efficient GPU miners can very nearly match those of you running FPGA miners

What are you basing this claim on? An FPGA board with a Spartan6 LX150-2 can do at least 150 Mhps and consumes less than 12 watts. That's at least 12.5 Mhps/Watt. And as far as I know even the most efficient GPU rigs haven't cracked 3 Mhps/Watt. I have a rig with four 5850s undervolted to 0.95 VDC, and I'm doing 1200 Mhps @ 450 watts from the wall. That's 2.67 Mhps/Watt, a rather efficient GPU rig, but not even close to the efficiency of FPGA. If I'm missing something here, please inform me.

[DeathAndTaxes]

You can do far better than that if you undervolt and underclock.  If you want efficiency try cutting the clock IN HALF and cutting voltage down 20%. Although you won't be getting to no 12MHps/Watt. 

Still comparing LX150-2 to an ENTIRE COMPUTER is silly.  LX150 will require a computer also.  Lastly I think the OP was indicating a GPU miner where electricity is cheap could underclock/undervolt and have a lower PRODUCTION COST (not raw efficiency) than a FPGA miner where electricity is expensive.

Given some parts of US have 6 cent electricity and some parts of Europe have 20+ cent electricity that is certainly possible.  The OP claim is on extreme and yours is the other extreme the reality is in the middle.

[ArtForz]

Huh? Heavily undervolted GPUs coming even *close* to FPGAs? Seriously?
So far the best I've seen from GPUs even with extreme undervolting is ~ 3Mh/J at the wall.
My first 32*LX150 box with a power-hungry mobo, WAY overkill fans and running at 1250mV Vccint? 6.27Gh at 382W at the wall, thats a bit > 16Mh/J.
A more optimized design, no overvolting, > 20Mh/J at the wall.
And S6s are real power hogs compared to higher-end chips, smallish V6 and Stratix IVs get >30Mh/J easily (at atrocious price/perf, but hey...).

For the math impaired, 20Mh/J at current difficulty -> 1BTC ~= 2kWh, that's $0.60/BTC at "european" $0.30/kWh and around $0.20/BTC at $0.10/kWh... yeah.

[Dargo]

You can do far better than that if you undervolt and underclock.  If you want efficiency try cutting the clock IN HALF and cutting voltage down 20%. Although you won't be getting to no 12MHps/Watt. 

Still comparing LX150-2 to an ENTIRE COMPUTER is silly.  LX150 will require a computer also.  Lastly I think the OP was indicating a GPU miner where electricity is cheap could underclock/undervolt and have a lower PRODUCTION COST (not raw efficiency) than a FPGA miner where electricity is expensive.

Given some parts of US have 6 cent electricity and some parts of Europe have 20+ cent electricity that is certainly possible.  The OP claim is on extreme and yours is the other extreme the reality is in the middle.

Fair enough. You're right that I wasn't figuring the power of the computer the FPGA boards are connected to (though I would think that this could be pretty low - a netbook perhaps). I wouldn't want to drop my voltage/clock as low as you suggest, but I see your point that by doing this I could come closer to the efficiency of FPGAs. This increases the $/Mh, but this ratio is high for FPGAs anyway. BTW, do you know anyone who mines at such a low voltage? 20% lower than what I've already got is 0.76 VDC. Don't think I've heard of anyone going that low. I'm not even sure I can. Anyway, don't mean to hijack the thread, but I'm interested in the efficiency issue, and wanted to know if there was any shred of truth to the OP's claim. I guess it was perhaps an overstatement, but some truth to it nonetheless.

[DeathAndTaxes]

Fair enough. You're right that I wasn't figuring the power of the computer the FPGA boards are connected to (though I would think that this could be pretty low - a netbook perhaps). I wouldn't want to drop my voltage/clock as low as you suggest, but I see your point that by doing this I could come closer to the efficiency of FPGAs. This increases the $/Mh, but this ratio is high for FPGAs anyway. BTW, do you know anyone who mines at such a low voltage?

No but very few people mine w/ FPGA either for the same reason.  Underclocking a GPU lowers the electrical cost per hash but raises the capital cost (same card produces less hashes).

Given the high risk in mining (who knows what bitcoin will be worth in 6, 12, 24 months) that tradeoff isn't currently worth it for most miners.  It is more profitable to run miners clocked to the max (and gulping power) as bitcoins are still selling for 3x power costs. 

Still for miners who have already bought hardware if necessary (due to falling bitcoin prices) underclocking would be a means to "stay in the game" longer if bitcoin prices fell significantly.

[Dargo]

Huh? Heavily undervolted GPUs coming even *close* to FPGAs? Seriously?
So far the best I've seen from GPUs even with extreme undervolting is ~ 3Mh/J at the wall.
My first 32*LX150 box with a power-hungry mobo, WAY overkill fans and running at 1250mV Vccint? 6.27Gh at 382W at the wall, thats a bit > 16Mh/J.
A more optimized design, no overvolting, > 20Mh/J at the wall.
And S6s are real power hogs compared to higher-end chips, smallish V6 and Stratix IVs get >30Mh/J easily (at atrocious price/perf, but hey...).

For the math impaired, 20Mh/J at current difficulty -> 1BTC ~= 2kWh, that's $0.60/BTC at "european" $0.30/kWh and around $0.20/BTC at $0.10/kWh... yeah.

Thanks Artfoz. So you are getting 196 Mhps and 12 watt per LX150 from the wall, and that's for your first non-optimized rig. No way are you going to touch that with a GPU rig obviously. Now if only FPGAs weren't so damed expensive.

[ArtForz]

Yep, and price/perf is where FPGAs fall flat on their face.
Why? Well, even at-cost a LX150 box still ends up at > $1/Mhps. And expected resale value is damn near 0.
With GPUs not only is the initial cost lower, but when the next generation comes around you can resell them for somewhere around 50-70% of the initial cost, keep the rest of the system and replace them with their successors.
If those end up about 40% better price/perf and perf/power (not totally unreasonable), you'd end up with the same hashrate as before but now using only 60% of the power. Repeat as often as necessary. Another benfit of that, with the short product cycles of GPUs, they'll be pretty much in warranty forever.
So... why even bother with FPGAs then, other than being low power? Well... density (32 LX150s in a 2U = trivial, 48 = still no big problem) and reliability/low maintenance (server grade PSU, 80mm fans with > 60kh expected lifetime, ...)
And of course having your own FPGA cluster to play with

[Raize]

I don't mean to cause controversy or anything, my point is that FPGA becomes a viable option easier for those of us paying $.06-.08/kWh than those of you paying $.25/kWh.

In summary: demand leads price leads difficulty

While this is true, supply decreases price significantly. If Bitcoins are worth $32/coin, the guy mining 1,000/month at $.08/kWh is doing far better than the guy mining 2,000/month at $.25kWh. My major point is that if Bitcoin price exceed greatly the cost to produce them, the price is GOING to decline, because the cost of entry into the market is too low. Do you seriously expect the price to stay at $32/coin if someone can hook up a 5850 and mine a coin in 3 days?

I mean, the only way something like this could remain supported was if all ATI cards were banned for export out of China or import to the US/Europe. But then you've got a black market on video cards, and I'd argue that would never happen.

Back on FPGA's, all the Europeans right now are doing is lowering the cost of entry for those of us in the US Midwest who enjoy very low power costs. Once the first major and profitable FPGA project starts taking orders, we can order thousands of units and again drive the price down to a slight premium over our costs.

[eldentyrell]

Still comparing LX150-2 to an ENTIRE COMPUTER is silly.  LX150 will require a computer also.

No, it just requires some device with an ethernet port that runs Linux.  You can use a crappy Linksys router as the "computer" driving your FPGA miners.  I use an Apple TV.

[grod]

Still comparing LX150-2 to an ENTIRE COMPUTER is silly.  LX150 will require a computer also.

No, it just requires some device with an ethernet port that runs Linux.  You can use a crappy Linksys router as the "computer" driving your FPGA miners.  I use an Apple TV.

Exactly this.  There's absolutely no need for an Intel CPU running Windows to drive mining boards.  None whatsoever.   Some cheapie routers (like my WNR3500 which is drawing ~8 watts from the wall) even have a USB port in case mining devices getting work via USB show up.

Oh, and my midwest power rate?  4.6c/kwhr + distribution = ~6c/kwhr for residential.  No doubt that can be lower if I get commercial rates.  The only question is: will bitcoin be around and profitable enough to recoup a hardware investment?

[memvola]

My major point is that if Bitcoin price exceed greatly the cost to produce them, the price is GOING to decline, because the cost of entry into the market is too low. Do you seriously expect the price to stay at $32/coin if someone can hook up a 5850 and mine a coin in 3 days?

If Bitcoin price exceeds the cost to produce them, difficulty goes up (precisely because of the people hooking up 5850's), so in your terms, the cost of entry is never "too" low. It can be low, or high, depending on how fast miners react. It's currently too high for instance, so the difficulty is going down.

EDIT: If you devise a competitive method to mine and keep it you yourself and find a big portion of the blocks for some time, then you could inevitably drive the price down. As long as there is open competition though, I don't see why it would happen.