electricity costs?

[salmi]

i am under the impression that electricity costs for mining btc are prohibitive

but i've just come across someone, someone who claims to know abit such thigs, and he says electricity costs should be as low a $1 or as high as $2.50 to mine a single bitcoin

i understand that costs will vary depending on where you are and how much retail electricity costs but is that estimate anywhere near reality?

please and thank you!

[1715411275]

1715411275

[1715411275]

1715411275

[1715411275]

1715411275

[1715411275]

1715411275

[1715411275]

1715411275

[1715411275]

1715411275

[Lauda]

It depends on your kwh/$ price and ASICs are very efficient when it comes to electricity consumption.

[DeathAndTaxes]

It depends on what hardware.

On a CPU, GPU, or FPGA they will consume more power than the BTC is worth.

However ASICs are a different story.  Looking at an "inefficient" 1st gen Avalon rig it uses ~ 8 J/GH.
Difficulty today is ~510 million.  So it takes on average 2^32 * 500M / 25 = 8.76 *10^16 hashes to create a bitcoin (25 BTC per block).  This assumes no stale shares, hardware errors, downtime, or pool fees.
At 8 J/GH that is 700,938,662 J or 194.7 kWh (3,600,000 J per kWh)

If we assume the average miner pays $0.10 per kWh, then using an Avalon rig would require ~$20.00 in electricity to produce 1 BTC.  That is just using the dificulty today, most people believe the difficulty will go much higher, probably to triple or quadruple the current value by the end of the year.  That would mean ~$80 per BTC using an Avalon (@ $0.10 per kWh).  The factors for the elecctrical cost of one BTC are efficiency (J/GH), electrical rates (USD per kWh), BTC exchange rate (USD per BTC), and difficulty (determines # of GH per BTC).  

With future not yet delivered hardware (highest 28nm specs) and $0.05 per kWh the cost may get close to $1 at current difficulty, however difficulty will be much higher by the time those devices are available (Dec/Jan).

[Lauda]

The next generation of ASICs should be even more efficient though.

[DeathAndTaxes]

The next generation of ASICs should be even more efficient though.

And difficulty will be even higher to compensate.  At 10x current difficulty a 0.8 J/GH 28nm chip will use the same amount of power per BTC as an Avalon does now.  The network is self adjusting that way.

[utarinues]

The next generation of ASICs should be even more efficient though.

Not much more power efficient.  28nm technology is probably best one can hope for even next year.

But If choosing if Avalon (or BFL from ebay) or KNC, I would go to KNC, it will be competetive much longer because more efficient

[Lauda]

The next generation of ASICs should be even more efficient though.

Not much more power efficient.  28nm technology is probably best one can hope for even next year.

But If choosing if Avalon (or BFL from ebay) or KNC, I would go to KNC, it will be competetive much longer because more efficient

Yes much more power efficient.
We already have 28nm? Huh?

[DeathAndTaxes]

OP might also find this interesting:
https://bitcointalk.org/index.php?topic=281279.0

[salmi]

thx, all. that's what i thought

the person who made the claim is francois velde, who happens to be a senior economist at the federal reserve bank of chicago

here, for your amusement, is my blog post about his howler

http://briansalmi.com/2013/11/11/chicago-fed-shoots-itself-in-the-head-over-bitcoin/

[ajax3592]

i am under the impression that electricity costs for mining btc are prohibitive

but i've just come across someone, someone who claims to know abit such thigs, and he says electricity costs should be as low a $1 or as high as $2.50 to mine a single bitcoin

i understand that costs will vary depending on where you are and how much retail electricity costs but is that estimate anywhere near reality?

please and thank you!

Electricity costs would depend on what "nm" is the fabrication of the chip of hardware that you'd be using. The lower the number the lesser the power consumption and vice-versa

[Rainkiss]

I'm still new to all of this bitcoin stuff, but I've been looking at maybe getting some cheaper equipment to do some basic mining cause it doesn't require constant attention.

While electricity costs are important, and I know faster is generally better, what numbers in terms of hashrate power would be recommended for a reasonable return? and something that would be relatively cheap. If I spend $100 on some USB miners and join a pool so that I'm getting steady progress rather than only getting paid if I find a new block, is it going to take a month or two to get my investment back, or is it going to take closer to a year while the difficulty continues to climb out of reach of newer participants?

I'm not looking for something that would make a lot, I'd be fine with gaining even a small profit.

[Gator-hex]

The next generation of ASICs should be even more efficient though.

Not much more power efficient.  28nm technology is probably best one can hope for even next year.

But If choosing if Avalon (or BFL from ebay) or KNC, I would go to KNC, it will be competetive much longer because more efficient

Yes much more power efficient.
We already have 28nm? Huh?

Yes 28nm is already in customer hands mining since Oct 2013 KNC won that race for Europe. A tad expensive but I'm sure they'll mine that back with a 2-4 month head start over competitors.

http://www.youtube.com/watch?v=qFjNUb1AbQ8

[Lauda]

So I was right. Next step will be lowering that to 22nm or less.

[DeathAndTaxes]

So I was right. Next step will be lowering that to 22nm or less.

Not anytime soon.  Also process node alone doesn't determine efficiency.  The 28nm chips by Cointerra and HashFast (just specs not yet delivered yet) have roughly 50% higher efficiency (~0.8 J/GH) than the first 28nm chip by KNC (~1.2J/GH).  Bitfury 55nm design is ~ 1 J/GH.  All these numbers are measured at the wall, companies often market the chip only efficiency but total power conusmption for a rig is higher.

Still the rate of improvement is going to slow dramatically.  In the span of less than year we went from GPU (330 J/GH) to high process node chips (~8 J/GH) and if Cointerra or HashFast deliver  before the end of the year we are looking at <0.8 J/GH.  

No foundry is offering 22nm (only lines at 22nm are internal fabs like Samsung and Intel) instead all the foundries (TSMC, Global Foundries, UMC) are going to 20nm and then 16nm.  However it is simply getting hard and harder to maintain Moore's law.  Economical 20nm chips are probably not going to happen until at least 2015 and 16nm seems doubtful before 2016 (IMHO 2017 seems more likely).  Neither provide the "massive" efficiency gains we have seen this year.  TSMC marketing (real world tends to be lower than marketing claims) on 20nm, is EITHER a 30% higher clock or 20% better efficiency at the same clock so if companies opt for the high performance route (faster chips) it wouldn't offer any efficiency improvement at all, and if they keeps clocks the same (no faster performance) we are still only looking at 20% lower power consumption.  16nm looks to offer up to 30% higher clocks AND 30% higher efficiency but it is even further out.

Summary
GPU ~330 J/GH  Jan 2013
First ASIC Miners ~8 J/GH March 2013
Best available today ~1 J/GH Aug 2013
28nm designs in production ~0.8 J/GH (spec) Dec 2013
20nm higher clock speed ~0.8 J/GH (est) ?2015
20nm same clock speed ~0.6 J/GH (est) ?2015
16nm in theory ~0.5 J/GH (est) ?2016-2017

As you can see the efficiency gains starts getting smaller pretty quick dropping real quick and the timelines start to get pushed out.  When you consider that in theory we might not have chips offering better than double current efficiency for more than a couple years and we went from 330 J/GH to 0.8 J/GH in a single year it kinda shows that in the short term how much of the potential efficiency has already been realized.  The prior poster didn't say it will never be better than 28nm simply that it is probably as good as it get for a while and it is a sentiment I agree with.

[Lauda]

So I was right. Next step will be lowering that to 22nm or less.

Not anytime soon.  Also process node alone doesn't determine efficiency.  The 28nm chips by Cointerra and HashFast (just specs not yet delivered yet) have roughly 50% higher efficiency (~0.8 J/GH) than the first 28nm chip by KNC (~1.2J/GH).  Bitfury 55nm design is ~ 1 J/GH.  All these numbers are measured at the wall, companies often market the chip only efficiency but total power conusmption for a rig is higher.

Still the rate of improvement is going to slow dramatically.  In the span of less than year we went from GPU (330 J/GH) to high process node chips (~8 J/GH) and if Cointerra or HashFast deliver  before the end of the year we are looking at <0.8 J/GH. 

No major foundry is offering 22nm instead going to 20nm and then 16nm.  Still economical 20nm chips are probably at least 2015 and 16nm seems doubtful before 2016 (or later).Neither provide a "massive" improvements to efficiency like we have seen int he past.  TSMC marketing on 20nm, is EITHER a 30% higher clock or 20% better efficiency so if companies opt for the high performance route (faster chips) it wouldn't offer any efficiency improvement at all.  16nm is looking like higher clocks AND 30% higher efficiency but at least on paper that is only ~0.5 to ~0.6 J/GH maybe.

Summary
GPU ~330 J/GH
First ASIC Miners ~8 J/GH
Best available today ~0.9 J/GH
28nm designs in production ~0.8 J/GH
20nm same clock speed ~0.6 J/GH
16nm in theory ~0.5 J/GH

As you can see the improved efficiency starts dropping real quick and the timelines start to get pushed out.  The prior poster didn't say it will never be better than 28nm simply that it is probably as good as it gets for a while.

Thank you for the constructive post. Good information can be acquired from it.
We will see how it goes in the upcoming years.