Bitcoin Forum

I'll just post the basic results here, the full analysis is here: http://organofcorti.blogspot.com/2013/04/913-asic-choices-asic-earnings-17-april.html

Previous post in series here (http://organofcorti.blogspot.com.au/search?q=%22ASIC+choices%22-canary).


http://2.bp.blogspot.com/-e4hV5dW9VIs/UW1k3AzL--I/AAAAAAAAHjk/S2kgeDipxAo/s1600/9.13.table1.2013-04-17.png (http://organofcorti.blogspot.com/2013/04/913-asic-choices-asic-earnings-17-april.html)

http://1.bp.blogspot.com/-9dBfC28ZZrc/UW1X1dnWSUI/AAAAAAAAHhs/aUa-JFxn2y0/s1600/9.13.chart2-2013-04-16.png (http://organofcorti.blogspot.com/2013/04/913-asic-choices-asic-earnings-17-april.html)

Are you considering the shipping of batches of Avalon ASICs in quantities of 10k to 200k chips per run available on demand to anyone? 
And with any tiny degree of accuracy can you extrapolate 24 months?

No, I don't have any definitive information about chip batch sales.


No. With the inaccurate shipping information available, even extrapolating 3 or 4 months in advance is going to be inaccurate. 24 months in advance, based on current shipment plans? Can't see how I could, sorry.

I do have a guess at how large the network will be once the first generation ASICs reach a sort of "steady state", and that's about 1000 Thps, with a pretty wide margin for error. But I have not idea if this would be achieved in 6, 12 months, 24 months or 36 months, just that it will be the next long term plateau that occurs.

Given that ASICs introduction was a step function game changer, you have a done a great job at modeling the near term.   For the longer term, I think you can only model that by backing into a difficulty where it is no longer that exciting to buy a miner because the ROI is pushed to the 9-12 month range and the return on selling the hardware is pushed down below a 50 GM.   Personally, I think that occurs around 1,800 TH but OBVIOUSLY price steps throw the monkey wrench in this.   Above 250 btc/usd, I think the equilibrium goes higher. 

I think it gets hard to determine ROIs at all - your costs are in fiat, but the cost of the device could be in denominated in BTC or fiat. From the work I've done, a network hashrate at next plateau could easily be 1800 Thps as you suggest. Do you have a time frame for this?

My model is just guess (trying to think of commodities and how people ALWAYS overshoot on capacity -- think chemical industry that even as an oligopoly they never stop increasing capacity until they are two steps too high -- look at TiO2 industry case studies for something that should be rational but goes through irrational capacity expansions).   So, that is just my BS caveat to say (I THINK people should stop at 1800 but I am pretty sure they blow throw it and people get burned and justify their "investment" as a new "hobby").
I think April of Next year.   And I think it happens in STEPS. 

Here ya go:
http://store.avalon-asics.com/?page_id=9605

Avalon ASIC Chip
The no bullshit, no fine print terms of sale
the only payment accepted is Bitcoin.
these chips will be available until at least end of the year, 2013.
spec and various information is available on the wiki
the chips being sold are packaged and tested.
the lead time on the chips is 9 to 10 weeks.
made to order from TSMC foundry and then packaged and shipped.
the minimum order quantity is 10,000 chips and the maximum order quantity is 200,000.
the chips are identical to those in Avalon, clocking 282Mh/s per chip.
communication protocol, reference board design provided in early May.
everything will be open source from FPGA to PCB design.
sample chips will be provided 4 weeks into ordering.
we do not offer technical support of any kind, this is final.
if you do not know what to do with the packaged chips, please do not purchase.
Please read the above carefully, as with all things Bitcoin one should treat this as an investment and make the decision best for you based on the liquid-able funds available at the moment when placing an order.

So is 1800Thps  and April the first step (plateau)? Or after several?

Yes, saw that. But I can't make any guesses as to how many they'll sell since I don't know if 200000 chips is the max per order or the max they'll sell ever, and I don't know what demand is like. It also isn't clear when units based on the chips will arrive because this depends not only on shipping times but also on r&d time for the bulk purchaser, and whether or not the bulk purchaser will be selling retail.

So I don't have sufficient information to even guess the number of chips that will be purchased or the time periods over which they'll appear. If Avalon release information regarding chip sales I'll try to add the chips in to the network and difficulty calculations, Estimates of the time when the chips start hashing would be inaccurate, but probably no more than the current estimates are due to BFL's delivery schedule.

I think every two weeks for the next three months we start see steps as difficulty adjusts and avalon plus ASICMiner constantly ship asics.   This gets extended (or the steps get larger) if UPS ever gets called to come to BFL.  I think we plateau next april after the people who get delivery of devices in december take months to realize they may never get their capital back......

BFL buyers from last year will never get their BTC back.

Mining equipment purchases are always speculation on the exchange price of BTC.  As such you always need to compare buying equipment against market purchases of BTC directly.

People also tend to substantially underestimate the work required to maintain a mining farm with reasonable uptime.

that is basically my point, people wade in hearing there are fish in the water but they do not stop to think if they should learn to swim until they are at least out 8 feet deep.   With a rip tide to boot.   they they complain they were not given a life vest.
that is why these things ALWAYS swing way past where they should have rationally stopped...

Maybe. But the idea is not to attempt plot the jumps in hashrate due to deliveries - I did that previously and tbh the inaccuracies in the estimations makes this no more accurate for my method. Instead, I'd use the estimates as a guide over time - I aim for the cumulative earnings per time period to be as accurate as possible rather than the hashrate to get there.

I agree that if the hashrate steps this will have a significant effect on the network hashrate and difficulty and possibly also cumulative earnings, but I don't think anyone can accurately estimate when these jumps will occur. As long as the hashrate shipping estimates are accurate, I think the method I use here is the simplest one that any miner can reproduce for themselves in order to plan for the future.

I'd be interested in seeing how you make your assessment though - I'm always keen to learn new things!

Insignificant initially but the ultimate impact will be HUGE.  
ASICMiner is shipping it's first 10 cards.   (https://bitcointalk.org/index.php?topic=178275.0)


Significant shipment of ASICS has begun.  
Avalon has sold more than 30,000 of their chip (https://bitcointalk.org/index.php?topic=177827.0) delivery in 10 weeks
10,000 more here (https://bitcointalk.org/index.php?topic=177994.0) delivery in 10 weeks
Generation 2 Bitcoin mining ASIC? (https://bitcointalk.org/index.php?topic=146371.0) delivery in ? weeks

Of course it will certainly take 10 weeks to build pcb's to mount them too, but let's assume the buyers know what they are doing and they have a plan to use their new toys.  Please tell me, oh great seer of seer's, what will the addition of 100,000 avalon chips (@282mh/s) do to the data and difficulty in 90 days.

Thanks for the update, viceroy - I appreciate you going to the trouble of collecting the data and putting it all in one place.

Do you have any clear information about the ASICMiner upgrade schedule? 200 THPS by when?

I can't believe it... but it looks like the first BFL's are out (https://www.facebook.com/CodingInMySleep). 

I do not, sir.

Wow. I can not believe this.. BFL Jalapeno unboxing and demo video oO.. Congratulations to this guy! Buuut.. "1 of 10.000" BFL's out there for customers :D

Hmm.. Difficult will rise in the next 4-6 weeks I think.

I've already taken into account 2Thps of BFL ASICs per day from May 15th. At worst, I may have to bring that forward a couple of weeks. I'll see if many more BFL devices start arriving before I change that date though.

What does the difficulty look like if it hits 1000THps by year end?

What about 5000THps?

You can approximate difficulty as follows:

Difficulty = network hashrate / 2^32 * 600

At 1000e12 hashes per sec, D ~ 140 000 000
At 5000e12 hashes per sec, D ~ 700 000 000


Of course it is really dependant on the previous retarget's 2016 block hashrate, but this at least will give you an idea.

These projections show how hurt your profit margin can become. For example:

Let's say you bought in on a FPGA farm a few months back or just did. A rig I was looking at was a $45000 investment, at 47.5GH/s using ~2200W of power.

So given electricity of .15/kWh, using today's $118.31 USD/BTC rate, and calculating for profit by day, you can see how it will affect your average profit per day given that the exchange rate stays the same, which it won't. I know it's not ASIC but it still shows you that even a lot of power will lose the ability to profit.


http://imageshack.us/a/img402/8689/captureoor.png


So given that kind of investment now or at least two months ago, you'd be looking at a while for ROI. And say you got your ASIC in August (a BFL Jalapeno for instance), with the difficulty level that high it would still take ~50 days to pay off the $274 investment.

All I can say is..ouch.

it looks like it could be even worse

Yep.

My point. That's a projection based off of a purchase from a month or two ago to today, that you're trying to pay off before the difficulty maximizes and the profit minimizes. I mean I like mining for fun but I do like to break even when it comes to electricity.

No, your projection is based on estimated data from now until August.

I said the purchase from a month ago or two today. The projection is from today until August.

Ah, I misunderstood. But if the purchase was from a month or two ago, wouldn't it be better to do your analysis from then as well?

Yeah, probably. See, I'm lazy and I didn't want to go looking down the old difficulty numbers, plus it works as a way of showing "if you bought today." And I meant "to today" and not "two today," damn touchscreens.

But I included all the old difficulty numbers in the initial post, you could use them. Anyway, you can cut short all your calculations by using the fifth column, which already provides the info you were calculating.

For example:

• If you bought on the 18th April the amount you'd earn until August is 2.897 btc / ghps -> 47.5* 2.897 btc = 137.6075 btc
• If you bought on the 1st February the amount you'd earn until August is 11.305 btc / ghps -> 47.5* 11.305 btc = 536.9875 btc

Hope that saves you some time, and I apologise for not making it more clear. Check out the blog post if you want more details.

New update in case you're interested: http://organofcorti.blogspot.com/2013/04/914-asic-earnings-23-april-2013.html


http://1.bp.blogspot.com/-MSoTAcfOy7k/UXZaDbe-PGI/AAAAAAAAHnQ/8KOk88tB9fI/s1600/9.14.table1.2013-04-23.png (http://organofcorti.blogspot.com/2013/04/914-asic-earnings-23-april-2013.html)

There was an interesting article on Wired posted on Reddit today, regarding the arms race of ASIC mining equipment.

http://www.wired.com/wiredenterprise/2013/04/bitcoin-mining-rigs/

.b

With increasing difficulty and reducing miners, which the price of 1 BTC in August? $ 500 +?

Not sure if already discussed (too lazy to read through the whole thread right now, sorry).
How about GPU miners that might move away when profit is close to 0 for them?

edit:
Nice work, an interesting read!

Bitcoins are created at a relatively constant rate (25 BTC every 10 minutes on average) that self-adjusts every 2016 blocks based on network hashrate. Rising difficulty and 'reducing miners' have no direct effect on supply/demand and, hence, price.

Well that shows you how much I'm paying attention. That extra month or two could have seriously helped if someone were going to invest that much to get ROI. Especially if they were on a loan.


The change in the price depends on the interest and the trading whereas the ability to mine profitably depends on the difficulty, market, electricity, etc.

Good question. I'm not including GPU miners leaving, since of the ~ 24Thps that were present prior to ASICs I don't know how much was attributable to GPUs and how much to FPGAs. So I don't know how much the hashrate would drop once GPUs are unprofitable. Another thing is that I've very carefully avoided guessing what the exchange rate will be, which is needed to determine GPU profitability.

All I can say is that some point we'll lose GPUs and at another we'll lose FPGAs. Many GPUs will probably leave early to mine LTC, FPGAs will mine on until the bitter end. I'm assuming that by the time this happens, 24Thps will be a very small fraction of the network, and probably won't have much of an impact on the forecast accuracy given all the other sources of error (shipping times being the largest source of error).

Just so.

Signus' statement implies that difficulty is affected by the USDBTC price and not visa versa. I have a number of posts on this subject (http://organofcorti.blogspot.com/2012/11/101-price-drives-difficulty.html), and a weekly forecast of the network hashrate and difficulty that does not rely on ASIC shipping predictions (http://organofcorti.blogspot.com.au/search?q=%22weekly+forecast%22), just previous hashrates and exchange rates. It is surprisingly (to me) accurate.

I said the ability to mine depends on the difficulty and many other factors.

I'm not saying the difficulty is necessarily affected unless the market implodes the market like it has done with ASICs. Essentially the price spike got people interested in bitcoin, and then because of that people are buying large amounts of mining power that escalates the difficulty rise. It's a cycle.

OK, well I'm going to have to disagree with you then. Even before the price spike there's been quite a nice relationship between the exchange rate and network hashrate. Check the posts I linked to.

Difficulty / network hashrate is predominantly affected by the exchange rate. If the things that the money buys are reduced or increased in cost, then then these external influences change the relationship away from expected.

I will do that, thanks.

At least an ASIC will pay itself off at a relatively decent rate given that most people will receive their ASIC devices once the difficulty is 45-55mil.

meh... if you look at capital expenditure (CAPEX) and operating expenditure (OPEX), ASICs are way better than GPUs.

For the same 2500USD spent, you'd now get a PC with 4x 7970 GPUs, giving you about 2.8GH/s for a total of 0.11 BTC/day (according to the table above, as of May 11th) and using up about 1400-1500W of energy.

Compare that to a 2500USD 50GH/s machine in August that will give you 0.40 BTC/day at 100W.

Even at 10.000USD per 50GH/s, serious miners won't loose money with respect to the current situation (if the table is correct)...

Sooo, my humble guess is... not much will happen.

I'm not following you. What post are you referring to?

@organofcorti: I was referring to "your" numbers table.

Long story short: I think ASIC mining will be where GPU mining now is, profit-wise, after the "ASIC-wall" hits.

The GH/s power required will simply have multiplied by a factor more or less equivalent to the price reduction in equipment of similar GH/s power. So, you will still invest the same amount for the same or slightly lower BTC return but you will also have cut your electricity costs.

The money-making window is NOW for those lucky few with a shiny new ASIC and it will close when the masses can finally get their hands on them too. Your bet is as good as mine as to when that will happen (given BFL should have shipped in 2012 already...).

Hope this clarifies the previous post a bit.

Ignoring electricity costs (which hopefully wont be an issue for a while unless we hit the petahash mark soon), is this what you mean:

http://4.bp.blogspot.com/-SKh5XgEniwo/UXphCbmIQ1I/AAAAAAAAHnw/RMBfHa2Ae90/s1600/9.15.chart1-2013-04-26.png (http://organofcorti.blogspot.com/2013/04/915-fair-price-for-asic-miner.html)

I would look at BTC/GH/s per $ invested.

As your projections show, BTC/GH/s is falling rapidly (nice blog BTW) but when you factor in CAPEX, the numbers look more mundane.

Now, to know whether you'll be making money in $ terms is another issue (as always, it depends on the market). My guesstimate is that the BTC should hover in the 50-70 USD range when all the crazyness has dissipated but things are very much in flux right now so I wouldn't bet on it ;)

At 50 USD/BTC, you're looking at a profit of circa 20 USD/day in August for 50GH/s (0.008 BTC/GH/s per day), not factoring power consumption. If you paid 50K for your ASIC (and got it in August), you just lost a ton of money. If you paid 2.5K, you'll get the same ROI as ppl used to get on GPUs when spending the same 2.5K on a PC.

I would plug in some CAPEX numbers in the tables/graphs to have a guesstimate as to how much one can invest and when, if one wants to make money (but it looks like you're doing it much better already :p).

I've included exchange rate in previous posts in the series, and produced tile plots of exchange rate / ROI with days to ROI as the tile colour. It's interesting, but at the moment everyone seems to want their estimates based in BTC, and no one seemed to know how to interpret the charts.

I agree it's an important factor, but back in January I'd only included exchange rates up to $40 and even then people though I was being optimistic. Point is, BTC market is too volatile to use an estimate - you have to produce a large number of estimates for a very wide range. If readers want that back, I'll bring it back next time.

Right, a projection on how fast they need to retire their GPUs.

I don't know how anyone could do that right now, given we have no clue how many ASICs are being put online and how many more are to come and when.

I should go and see a psychic, then I will let you know :)

I can't make heads or tails of those graphs - can you explain what they show please?

Sorry I'm at work right now and don't have time, but if you click on the chart the blog post will open which explains it all in detail.

Somehow, I don't think it will clarify things :)

Litecoin has been consistently ~1.3-1.5x more profitable than BTC for GPUs for months now.  I expect most GPUs have already moved.

(Indeed, a large % of LTC GPU miners were even switching between alt coins because they were more profitable still, but nowhere near as stable as LTC)

if one orders 10 000 avalon chips today and they deliver as they say within 12 weeks
they should be shipped around mid august.

let´s say getting them mounted on boards and running will take 3 to 4 weeks = mid september

that will give you 2800 GH/s with an investment of...(based on current btc rate)

(90 k $ for the chips, + 70 k $ for assembly =) 160 k $

this is massive investment in the best available mining hardware. with a difficulty of 80 mio it would generate 530 btc per month. so it would need about 3 months to pay back.

send cash, i order !  ;D

Time to sell you house :)

my wife wont let me, but i am working on it  ;)

Sell your wife AND the house!

Problem solved ;D