ASIC Difficulty Curves

[bitboyben]

So I figured I'd break down the difficulty chart in to two. One near term which is driven by expected ASIC deliveries and the longer term difficulty increases which will be driven by ROI.
I figured I'd get the ASIC deliveries part of the difficulty curve out of the way because it seems more predictable and less controversial as most people's models have fallen some where within reasonable limits of these ranges.

The goal was to give some measuring tools for miners to evaluate the rates of change to the difficulty such that they can have plan A, B and C for low, medium and high levels of difficulty rather than say this is the most accurate model base your numbers on this (little shout out to the book "The Flaw of Averages" for that idea).
There may be limited application to rising or falling exchange rate as well.  Increasing values of BTC is as good or better than lowering the difficulty.
The core question being answered is "Given my situation: Do I reinvest in hardware and if so how much?" Of course that is a personal question and your individual variables and values may be quite different from mine.

What is in there?
First off I decided to stick with TH on total network hash rate as the measure of difficulty.
BFL 150TH, bASIC 50TH, Avalon 40TH
6,8,12 weeks roll outs.
240TH, 250TH, and 270TH
240 b/c total of deliveries
250 b/c total of deliveries plus some reserve GPU/FPGA users hanging on and a small bonus in case released hardware runs slightly better than advertised.
270 b/c total of deliveries plus approximately the current max TH I've seen on some of the Total hash rate charts.

As the graph approaches the estimated delivery amounts I tag on a simple +2% per time period.
I did this for denoting there will be a continual increase after the deliveries are out the door.
I don't plan on using this for the longer term difficulty increases.

I gave GPU Users till about 70TH until they would drop out.
As for FPGA users I guessed most would upgrade if they could as there are many good avenues for that.
So of the 25TH attributed to those groups I dropped it to 5TH for those hanging on for various reasons above 70TH.

Used logistic function as the basis for the chart.
I'm going to carry on with some more scenarios based on the population growth models with the follow up longer term difficulty chart.
I think the parallels between BTC mining and Population growth in ecology are a great match.
http://en.wikipedia.org/wiki/Logistic_function

With that in mind I think the *quicker* the difficulty rises the *slower* the long term difficulty will creep towards theoretical maximums.

So far this is more pessimistic than my original numbers over the short term.

[1715187908]

1715187908

[1715187908]

1715187908

[organofcorti]

I think you've detailed your estimate of the initial hashrate increase quite well. However estimating a longer term difficulty change will require an estimate of the USD-BTC exchange rate since the exchange rate seems to drive difficulty, with a slight lag. If you can derive an equation to relate the two historical values, it might lend your model greater verisimilitude.

[MrTeal]

I think you've detailed your estimate of the initial hashrate increase quite well. However estimating a longer term difficulty change will require an estimate of the USD-BTC exchange rate since the exchange rate seems to drive difficulty, with a slight lag. If you can derive an equation to relate the two historical values, it might lend your model greater verisimilitude.

That the big problem with estimating longterm difficulty, since anyone who claims to know what the price will do is either filthy rich or lying. Even with bitboyen's estimate of 270TH/s giving a difficulty of ~38M, very few people are going to invest in new ASICs if the price drops down to $5 without massive hardware price drops. It will be interesting to see what the mining factor 100 (or maybe USD/24hr@100GH/s now) will be once things stabilize in the spring.

[organofcorti]

......

Even with bitboyen's estimate of 270TH/s giving a difficulty of ~38M, very few people are going to invest in new ASICs if the price drops down to $5 without massive hardware price drops. It will be interesting to see what the mining factor 100 (or maybe USD/24hr@100GH/s now) will be once things stabilize in the spring.

If bitboyben's calculations are correct, D=38 million would be due to already purchased ASICs - unless the exchange rate drops very low they will be running constantly to get a return on their investment sooner. No moer ASICs need be purchased.

If the rate does drop significantly (like your US$5/BTC suggestion) D will only drop slowly - people paying more electricity will stop mining first.

[jl2035]

I will be very interesting when all that hw gets shipped..

[bitboyben]

I would like to suggest using $10 as the price when calculating difficulty curves as it is close to the current price first off and secondly of the price changes those charts can be more easily adjusted percentage wise.

38million difficulty Have we already peaked? And so soon? This can't be good for hardware sales.

[DeathAndTaxes]

In the 3 months between May and August 2011, hashrate increased by a factor of 10 as everyone started GPU mining. It was then basically flatlined until for a year until this summer when the price jumped. Given the assumption that people are making about price staying constant, I'd say it's much more likely that we'll see an initial huge surge as all of 6-12 month payoff slack is taken up, and then a tapering off of hashrate growth.

One thing to keep in mind is that ASICs introduce a new variable... hardware pricing.   While they have large upfront costs the per unit cost is far lower than current retail price.  So if/when new sales flatline (and difficulty stagnates) ASIC sellers can drop prices significantly.   Sell the same unit for 50% of current price and suddenly it looks very profitable and sales start to pour back in (and in time difficulty takes another leg up).  Months later the same stagnation occurs and one can cut prices again and again and again.  After selling the same units at 3 or 4 price points the profits can be rolled into an improved design at a smaller manufacturing process and the same process starts all over.

We didn't see that "much" with GPU.  Yeah as 5000 series hit end of life there were some "sales" and closeouts which resulted in lower hardware cost (MH/$) and then later buying used GPU lowered it further but the frequency of such price revisions and the magnitude was much less than you can expect from ASICs.

[MrTeal]

In the 3 months between May and August 2011, hashrate increased by a factor of 10 as everyone started GPU mining. It was then basically flatlined until for a year until this summer when the price jumped. Given the assumption that people are making about price staying constant, I'd say it's much more likely that we'll see an initial huge surge as all of 6-12 month payoff slack is taken up, and then a tapering off of hashrate growth.

One thing to keep in mind is that ASICs introduce a new variable... hardware pricing.   While they have large upfront costs the per unit cost is far lower than current retail price.  So if/when new sales flatline (and difficulty stagnates) ASIC sellers can drop prices significantly.   Sell the same unit for 50% of current price and suddenly it looks very profitable and sales start to pour back in (and in time difficulty takes another leg up).  Months later the same stagnation occurs and one can cut prices again and again and again.  After selling the same units at 3 or 4 price points the profits can be rolled into an improved design at a smaller manufacturing process and the same process starts all over.

We didn't see that "much" with GPU.  Yeah as 5000 series hit end of life there were some "sales" and closeouts which resulted in lower hardware cost (MH/$) and then later buying used GPU lowered it further but the frequency of such price revisions and the magnitude was much less than you can expect from ASICs.

True, though there will be limits to that. BFL is obviously in the drivers seat there, if they can get their costs down. I'm unsure how low the price on something like ASICMINER could go, given their 4.2GH/J efficiency. BFL might be able to get 1TH/s down to $5000, but ASICMINER would have trouble with that when they have to deal with powering, housing and cooling 4.2kW.

That's all assuming price stays constant, as I said. If (and when) prices for hardware drop (or the price of BTC rises), you would expect another inrush of hashing power.

[bcpokey]

In the 3 months between May and August 2011, hashrate increased by a factor of 10 as everyone started GPU mining. It was then basically flatlined until for a year until this summer when the price jumped. Given the assumption that people are making about price staying constant, I'd say it's much more likely that we'll see an initial huge surge as all of 6-12 month payoff slack is taken up, and then a tapering off of hashrate growth.

One thing to keep in mind is that ASICs introduce a new variable... hardware pricing.   While they have large upfront costs the per unit cost is far lower than current retail price.  So if/when new sales flatline (and difficulty stagnates) ASIC sellers can drop prices significantly.   Sell the same unit for 50% of current price and suddenly it looks very profitable and sales start to pour back in (and in time difficulty takes another leg up).  Months later the same stagnation occurs and one can cut prices again and again and again.  After selling the same units at 3 or 4 price points the profits can be rolled into an improved design at a smaller manufacturing process and the same process starts all over.

We didn't see that "much" with GPU.  Yeah as 5000 series hit end of life there were some "sales" and closeouts which resulted in lower hardware cost (MH/$) and then later buying used GPU lowered it further but the frequency of such price revisions and the magnitude was much less than you can expect from ASICs.

That's a very good point, that I hadn't really considered previously, so thanks for bringing it up. It's actually an interesting distortion to the curve, a cost-bending to hardware to combat the almost inevitable hardware sale stagnation. One can only hope that it would be done in a timely fashion, as too immediate a drop would be tantamount to spitting in the early-adopters faces ("thanks for bankrolling development, now all the johnny come-lately folks get a free ride on your buck, and knock down your ROI").

If companies take the long view though, and BTC remains relatively stable/successful, I can see this leading to good things as the new tech becomes more accessible to all, and of course, as stronger hash rates shore up the security of bitcoin.

[organofcorti]

Used logistic function as the basis for the chart.
I'm going to carry on with some more scenarios based on the population growth models with the follow up longer term difficulty chart.
I think the parallels between BTC mining and Population growth in ecology are a great match.
http://en.wikipedia.org/wiki/Logistic_function

Try using the log-logistic or log-normal functions. I think they might be a better model and more what you're after.

[bitboyben]

Used logistic function as the basis for the chart.
I'm going to carry on with some more scenarios based on the population growth models with the follow up longer term difficulty chart.
I think the parallels between BTC mining and Population growth in ecology are a great match.
http://en.wikipedia.org/wiki/Logistic_function

Try using the log-logistic or log-normal functions. I think they might be a better model and more what you're after.

Ok I'll try those out as well. I guess my decision to use population growth is a bit political in nature. I think lots of miners are blissfully unaware of how things are going to go down. BTC per day is a limited resource and we are all competing for finite resources.

[organofcorti]

Used logistic function as the basis for the chart.
I'm going to carry on with some more scenarios based on the population growth models with the follow up longer term difficulty chart.
I think the parallels between BTC mining and Population growth in ecology are a great match.
http://en.wikipedia.org/wiki/Logistic_function

Try using the log-logistic or log-normal functions. I think they might be a better model and more what you're after.

Ok I'll try those out as well. I guess my decision to use population growth is a bit political in nature. I think lots of miners are blissfully unaware of how things are going to go down. BTC per day is a limited resource and we are all competing for finite resources.

But we don't reproduce. With our technology I mean. Well, I never did, anyway. Maybe others are not so squeamish

[ercolinux]

......

Even with bitboyen's estimate of 270TH/s giving a difficulty of ~38M, very few people are going to invest in new ASICs if the price drops down to $5 without massive hardware price drops. It will be interesting to see what the mining factor 100 (or maybe USD/24hr@100GH/s now) will be once things stabilize in the spring.

If bitboyben's calculations are correct, D=38 million would be due to already purchased ASICs - unless the exchange rate drops very low they will be running constantly to get a return on their investment sooner. No moer ASICs need be purchased.

If the rate does drop significantly (like your US$5/BTC suggestion) D will only drop slowly - people paying more electricity will stop mining first.

Due to nature of ASIC people that mine with them will not stop unless they lose lot of money in doing that (2 or 3BTC of power for every BTC mined IMHO): no way to resell them plus the fact that probably new generation of even faster or less power hungry ASIC can be presented in any future so in case of price drop keep the ASIC offline it's a direct loss of hundreds of $, it's less dangerous lost some money in mining and keep the bitcoin hoping that in future price will rise again.

[kjlimo]

Used logistic function as the basis for the chart.
I'm going to carry on with some more scenarios based on the population growth models with the follow up longer term difficulty chart.
I think the parallels between BTC mining and Population growth in ecology are a great match.
http://en.wikipedia.org/wiki/Logistic_function

Try using the log-logistic or log-normal functions. I think they might be a better model and more what you're after.

Ok I'll try those out as well. I guess my decision to use population growth is a bit political in nature. I think lots of miners are blissfully unaware of how things are going to go down. BTC per day is a limited resource and we are all competing for finite resources.

What?  BTC per day is a limited resource?  How are things going to go down?  Oh knows... I'm so blissfully unaware....?!?!?!?

[MrTeal]

Used logistic function as the basis for the chart.
I'm going to carry on with some more scenarios based on the population growth models with the follow up longer term difficulty chart.
I think the parallels between BTC mining and Population growth in ecology are a great match.
http://en.wikipedia.org/wiki/Logistic_function

Try using the log-logistic or log-normal functions. I think they might be a better model and more what you're after.

Ok I'll try those out as well. I guess my decision to use population growth is a bit political in nature. I think lots of miners are blissfully unaware of how things are going to go down. BTC per day is a limited resource and we are all competing for finite resources.

What?  BTC per day is a limited resource?  How are things going to go down?  Oh knows... I'm so blissfully unaware....?!?!?!?

Maybe for you. I bought a few ASICs and now I'm going to retire.

[Joshwaa]

I ordered 5 and I have already retired. This is good no?

[bitboyben]

Well BFL just announced that they will be receiving 100,000 chips. Does this mean ASIC mining will become marginally profitable in mere months?

[ercolinux]

Well BFL just announced that they will be receiving 100,000 chips. Does this mean ASIC mining will become marginally profitable in mere months?

100.000 chips are 12500 Single boards or 500 MiniRig or a total of 750.000GHash/s -> 750THash/s. If is true this make ASIC not profitables at all unless a big jump in bitcoin price: if only BFL sell ASIC with this numbers a Single will take 2 years to repay itself at $0,15/KWh and a MiniRig 500 days.

[MrTeal]

Well BFL just announced that they will be receiving 100,000 chips. Does this mean ASIC mining will become marginally profitable in mere months?

12,500 Singles at current prices is over $16M worth of product. Do you think BFL is going to be able to sell that much in a few months?

[bitboyben]

Yes good point.
I guess it means that they can sell it as fast as they can make them.
If they can push out 20,000 chips in ~ two months then it seems possible to get the  other 80,000 in eight months.

But yes sales are also key.

If I was to take my assumptions further it would put pressure on miners to buy more hardware sooner while difficulty is projected to be lower.
But that would increase the Sales/difficulty even faster.

Thoughts?