Are you high?
$220 per GH/s?

Even if Bitcoins hashrate was flat (~70 mill) and never rose Ixcoin difficulty would rise if more people merged mined it.

13/70 = ~18% of Bitcoin hashpower is merge mining Ixcoin.  That leaves a lot of room for difficulty to rise.

Was this on the reduced speed boards?  The reason I ask is if the boards intended for 25 GH/s are getting 21 GH/s then they are essentially underclocked (even if unwillingly) and that will improve efficiency. 
Was the 116W increase at the wall? 

The next process node would be 22/20nm.  However costs today are much higher per transistor than 28nm.  That combined with a new NRE would make 22/20nm miners uneconomical today.    To date only Intel has released product using 22nm process and Intel doesn't use foundries they built their own custom fabs to their own specs.  It doesn't look like 2014 will change that much.  A couple companies are looking to do some production runs of cellphone chips @ 20nm.  Remember cellphone is a perfect market for this as the higher chip cost isn't really that important but improved battery life is.  General purpose fabrication @ 20nm is probably not going to happen until 2015 and any miner you buy today will either have made a profit or never will long before that.

Things you should be thinking about:
a) Bifury, Avalon, and ASICMiner will move to 28nm eventually.  Bitfury 55nm chip has better efficiency than some 28nm designs.  What will its 28nm design bring.
b) Miners are bad at math.  They may collectively buy hundreds of PH/s of capacity.  So much capacity that collectively miners lose (total hardware cost + total electrical cost < total mining revenue).
c) Lots of room for falling prices @ 28nm.  Raw silicon is likely $0.25 per GH.  The material cost for an entire rig is likely $1 per GH.  That doesn't mean you will see $1 rigs but prices can come down a lot.  

"b" is the one to worry about.  You can't control it so your potential fortunes (or losses) are highly dependent on the actions of someone else who may be utterly clueless. 

If you are still worried about 20nm, remember just because a process node exists doesn't mean it is the cheapest way to produce a chip.  Mining is very capital intensive.  Miners are constantly looking for falling prices not rising prices.  It generally takes two to three years before a new process node becomes cheaper than the prior one.  Worse this process seems to be "slowing down".  The cost improvement when mature from one generation to the next is getting smaller and the time it takes to mature is getting longer.   So it takes more and more years to get a smaller and smaller gain.  These slides are from a presentation where NVidia was "complaining" about TSMC's 28nm process last year but the the general theme is applicable to 20nm and beyond.  Even NVidia (with annual volume in tens of millions of units) doesn't project their cost per transistor on 20nm to beat 28nm until Q1 2015 and 14nm to beat 20nm until Q1 2017.  

Oops.  Fixed.

Congratulations John (and rest of HF team).  Big milestone.  I look forward to the joint press release. 
I assume this means you still anticipate meeting your delivery estimate?

Lowered efficiency of Avalon from 8.5 J/GH to 8.8 J/GH.  The Avalon miner is shipped in multiple configurations with 2, 3, or 4 modules in three different batches, using different power supplies.  The goal of the table is to provide an average break even point.  Reported efficiency "at the wall" ranges from 8.5 J/GH to 9.3 J/GH. 

I don't know if there ever was a v1.2 released publicly; this is the only photo of the PDU I have seen.

I get that the capacitor smooths out voltage but this board is being connected to a highly regulated power supply.  Are the tiny capacitors here having any meaningful effect compared to the PSU itself.  ATX PSU should be able to handle pretty significant changes in current and keep the voltage stable.

Anyone know if there is a board diagram for the PDU available?

Always wondered.
Looking at the photo it appears the caps are for the 12V distribution and the ATX PSU provide very good regulation why does the PDU need caps?

The short side of the curve looks probable.  In the pre-order guesstimate threads we are looking @ 6 PH/s of preorders by end of December and that is assuming KNC is only 0.5 PH/s.  KNC has been one of the ones which it is difficult to estimate in aggregate because there is no die information (to estimate wafer batch size), no pre-release on total amounts, etc. 

That is why we are looking at pre-orders.  A large portion of the initial 6 PH/s in 2013 has ALREADY been paid.  Now maybe miners overpaid and will never see a return but that hashing power will come online regardless because it is a sunk cost.   Take Avalon chips as an example.  There is a very good chance those that bought them will never see a positive return because they aren't online yet however once manufactured they are a sunk cost.  Not mining just means a -100% ROI.  Mining means something better.  They may still lose but a -20% ROI is better than -100% so they will be used ... by somebody.

Now I agree with some of the numbers it is tough to estimate how many sales have occured.  Cointerra indicated they are selling 2 PH/s.  How much of that 2 PH/s have they sold.  The good news is most of the "may not be able to sell" are further out.  I would consider the 2013 numbers to be of a higher confidence.  I would also point out that the amount of capital depends on the selling price.  The cost per GH (excluding NRE) of 28nm silicon is probably less than $0.25.  An entire system is probably about $1 (people forget how much power supplies including DC to DC PSU cost).

So if we assume 6 PH/s has already been paid the question becomes how much more capacity could be sold at prices declining to $2 per GH/s.  

Simple version:
Price reflects supply and demand not cost.  If people are buying $20 per GH/s and you are selling out why would you lower the price.  You lower the price when it is higher than your marginal cost of production AND you are no longer able to sustain sales at the current price.  I think we can get to 100 PH/s eventually just by cutting prices however that is harder to project.

Oops.  It was late.  I misread 0.11 BTC per GH as 0.11 BTC per chip.  Still I don't see late 130nm as a viable option.  There is nothing wrong with 130nm in June 2013.  Look how quickly it allowed Avalon & ASICMiner to get "to market".  130nm in November?  I don't see it being attractive not when there is cheaper more efficient silicon.

Cointerra beats Monarch in every respect.  Delivery date, price, and efficiency.  Not saying one should buy it trying to project ROI% for Dec with some much hashpower being sold (consensus guestimate - >6 TH by end of year).  However there is no reason anyone should pick Monarch over Cointerra. 

In the long run I think the successful ones will.   Look at Bitfury.  They sell no boards to the public instead they wholesale chips and rely on three major OEM (US, EU, Russia).  Think of how well that allows them to focus on the core task of chips, chips, chips.  Given their 55nm design has higher efficiency (J/GH and GH/mm2) than some 28nm designs it seems to be working.  If I was a ASIC company I would be worried about what a 28nm die shrink will look like.

Even AMD doesn't make graphics cards.  They make GPUs for a dozen or so OEMs which make graphics cards.

Locking.

https://bitcointalk.org/index.php?topic=283820

Reserved.

Reserved for highlights from original thread.

Moderated version. The original thread (now locked) is useful to see how the numbers evolved.

The idea
The purpose is to try to guesstimate (and yes that means error) the amount of hashpower in "the pipeline".  If there were no pre-orders, difficulty would rise based on sales and that makes future sales less attractive (lower ROI%) which reduces the amount of future sales.  This economic feedback model constrains hashing power based on efficiency, electrical cost, capital cost, exchange rate, the risk premium miners are willing to accept and the time value of money.    GPU didn't continue to grow to petahash scale because as new hardware was deployed it lowered the reward and increased the risk on future hardware and this slowed the growth until a rising exchange rate improved the reward.

Preorders break this feedback cycle as difficulty doesn't rise until delivery.  This can result in sold (but undelivered) capacity growing faster and larger than would happen if it was constrained by immediate feedback.  This is a great scenario for hardware vendors; they can sell more hardware, earlier and at higher prices.  On the other hand it means miners may face an avalanche of hardware purchased by their peers who don't quite understand how high difficulty can go. Accurately projecting growth for any period of time longer than a month is probably an exercise in futility.   We can however look at it from a different direction; if we can guesstimate the amount of preorders and can also estimate delivery timeframe then we can use that for the basis of a curve. So throw me your cites, guesstimates, and official numbers to try an pin down how much capacity has already been purchased.  

The rules
1) To keep this thread from derailing please leave the "xyz is a scam" or "abc sucks because ..." posts for another topic.
2 ) No idea/number is bad.  Please be respectful.  Honestly nobody knows for sure, except the chip makers and they aren't talking.
3 ) To stay on topic I will erase anything even remotely in violation of rule #1 and #2
4 ) A good starting point is the total amount preordered.  It may not be realistic but it provides a starting point.
5 ) Right now we aren't so much worried how likely a plan is just that there is a plan.  We can discount the hashpower by probability later (on edit: starting separating unfunded plans from presales and investor funded operations)
6 ) If you have a reference or cite (even unofficial) to back up a guestimate please link to it.  
7 ) If there is no reference a reasonable explanation is more useful then just posting a number. "Show your work".
8 ) We can safely assume that all non-ASIC hashrate will go to zero so no need to break it out between delivered and ordered.  Eventually hashrate ~= total pre-orders.
9 ) Mining companies buying retail product won't be listed seperately as they should already be included in the chip makers total.  Large operations should be noted so we can improve the estimate for the chip maker.

The running total

Promised delivery by December 2013

Announced product but insufficient information to make an informed analysis (should be considered lower confidence and conservative estimators may wish to exclude completely)


Post 2013 rollouts






Relationship between difficulty and hashing power

Upper limits on difficulty based on hardware efficiency:
Miners are unlikely to mine when their electrical costs are higher than the value of BTC mined.  This limit can be called the electrical break even point and is based on:
a) the current exchange rate (USD per BTC)
b) the hardware efficiency (J/GH )
c) the miner's electrical rate (USD per kWh)

When hashrate/difficulty gets high enough it will cause the least efficiency miners to idle thus creating a sort of replacement cycle (i.e. x GH/s new efficiency hardware causes Y GH/s of older less efficient hardware to idle).  This should slow growth significantly because the returns on new hardware will be low, miners will be exposed to the bad news of less efficient miners being forced to idle and X GH/s doesn't mean the hashrate only rises by (X-Y)/GH.  It also illustrates the improbability of difficulty power growing exponentially over a long period of time like a year.  For example 65 million difficulty gaining 75% per month for a year results in 50 billion difficulty.  The electrical cost even at 1W/GH and $0.10 per kWh would >$250 per BTC.  

A related thread on the break even point is here:  Break even difficulty by hardware efficiency (power cost = value of BTC)






[1] https://docs.google.com/a/nacrypto.com/spreadsheet/ccc?key=0AiLYkKIHJaIsdHpIaGdUOWRYVUdncTNpNlVKbVhCbEE#gid=0  970,000 chips @ 282 MH nominal
[2] 550 orders @ 400 MH nominal
[3] Based on report that Dave (US distributor) sold out of their allocation of 300 full systems and 300 starter systems.  Oct is not sold out but conservatively it will if/when Aug deliveries are made.  I will assume that the EU distributor received an equal allocation (an assumption based on bitfury facing unknown demand and users in both markets).
[4] http://bitcoin.stackexchange.com/questions/8577/how-much-asic-power-has-been-or-is-being-shipped-in-2013  Crude assumption based on distribution of wait list (hashing power per order) and number of orders.  2PH/s is guestimated based on (avg GH/s per order of known orders)*(num order numbers)*(1/3 to account for unpaid/test orders).  Monarch is highly unlikely to ship in volume (if at all) in 2013 while upgrades cancel the existing 65nm order which would reduce the amount of 65nm pre-orders.
[5] Guestimate. https://bitcointalk.org/index.php?topic=278384.msg2994436#msg2994436
[6] https://bitfunder.com/asset/IceDrill.ASIC
[7] http://www.dpcapital.net/blockchain/?hours=336
[8] http://en.wikipedia.org/wiki/Occam's_razor
[9] Reduced to 30 full systems in Aug for both US and EU distributors https://bitcointalk.org/index.php?topic=278384.msg3010364#msg3010364
[10] http://www.coindesk.com/cointerra-cuts-price-of-terraminer-iv-bitcoin-mining-rig/ "In Dec" without a specific date can mean as late as 31 DEC.  Given that and the tight schedule and the fact that even a small delay would push it into 2014 I included it in the 2014 group.  Cointerra promises 20% additional hashrate if more than 30 days late so bump the amount from 2 PH/s to 2.5 PH/s.
[11] https://ghash.io/
[12] HashFast MPP will issue miners up to 4x their initial hashing power if 100% ROI is not acheived within 90 days.  Even if MPP is not needed, HashFast would need the chips in reserve and any chips not paid out in the MPP are likely to be deployed as quickly as possible.
[13] http://www.coindesk.com/bitmine-to-drop-4phs-of-asic-power-onto-bitcoin-network-before-april/
[14] http://thegenesisblock.com/cointerra-expects-to-deliver-2-phs-of-asics-in-december/ (Numerous references, Cointerra 2PH/s, KNC 0.5 to 2 PH/s, AsicMiner 1 PH/s, Bitfury 0.5 PH/s, Avalon 0.32 PH/s
[15] Upgraded BFL estimate from 2 PH/s to 3 PH/s https://bitcointalk.org/index.php?topic=278384.msg3029092 (see also the next two posts)
[16] https://picostocks.com/docs/index/19
[17] https://bitcointalk.org/index.php?topic=264696.msg3033042#msg3033042
[18] https://bitcointalk.org/index.php?topic=283820.msg3046590#msg3046590
[19] http://thegenesisblock.com/easic-announces-24-ths-hash-fast-miner-activemining-shares-move-70/
[20] https://bitcointalk.org/index.php?topic=283820.msg3125451#msg3125451

You can move it yourself.

Lower left of this page [Move Topic]

Per the break even thread ...

@ muyuu,

I think you misread (or I wasn't clear).  The scenario is the network is in equilibrium (hashrate growth changes with exchange rate because the difficulty is near the collective break even point for the network).  To illustrate lets say the exchange rate is $200 per BTC and there is a miner with an older 130nm rig which requires $180 in electricity to produce 1 BTC (based on difficulty, efficiency, and power cost).  Now the exchange rate falls and remains below $150 for thirty days.  At some point that miner is going to stop producing one BTC worth <$150 using >$180 in electricity and just buy BTC because it is cheaper.  People may make bad long term plans but generally make better short term ones.  When you are spending >1 BTC to produce 1 BTC it is pretty easy to see you should stop.  On the other hand he is unlikely to throw his rig away.  If the price (or more importantly price/difficulty) rebounds then we will turn his rig on.  The bad news for this marginal miner is that the higher price/difficult will bring in new investment likely by miners with lower power cost and more efficient gear so difficulty will likely overshoot his break even point and he will have to idle again. 

We aren't concerned with hardware costs for the break even point as you indicate they are a sunk cost even if a miner has a lifetime loss it still makes sense to mine to reduce that loss UNTIL difficulty exceeds the break even point.  At that point any additional mining is just increasing the loss for the miner.  

We saw this in dynamic in the GPU era where difficulty would track price.  Today it doesn't simply because we are no where near the break even point for the network.  For even the least efficient gear and the highest power cost user it would be >6PH/s.