Topic: BFL ASIC & Difficulty Profile (Read 5614 times)

For any GPU miners concerned about the speculated increase in difficulty hurting their earnings (especially in combination with block reward halving), the new Coinlab pool for GPU miners that guarantees minimum earnings might be of interest to you.  Here is a link to the thread with a run through of the program https://bitcointalksearch.org/topic/ann-coinlab-protected-pool-99643

I just ordered, and the order number was in the mid 5000's         

Keeping it simple as I can.
IMHO

https://bitcointalksearch.org/topic/wait-list-bfl-sc-pre-order-information-monarch-orders-too-89685

Approx 2500 orders there will be more orders as time goes on of course but then every order number doesn't = an actual unit.

I just assumed the SC single per order = 40Ghash

So I'm gonna give it a range

1250-3000 orders

50 Thash -> 120Thash  4-10X difficulty

Personally I think most GPUs will become unprofitable at the lower end of that threshold and I'm not sure how many FPGAs are out there so I'd say -80% of the current hash rate based on GPUs will dissolve. So +/- on that by 20%

As for Shipping times: BFL gave themselves till OCT +2 months and according to our records they are about two weeks slower than they say 4-6 weeks = 6-8 weeks.

I'd say the very first shipments for the first month of orders will start shipping last week of Sept until as late as the start of the new year. (Oct + 2 months plus "awe screw it it's only two more weeks late" factor)
There's roughly 2000 orders for the first month that technically need to be shipped by the end of that period. But If they get the first batch out the door close to their stated targets then would the guys further down the order list really ask for their money back at that point?

So I'm thinking
4-10X difficulty increase (+/- 20% Hash) between Oct 2012 and Feb 2013

But what happens over the following Six months should be interesting...
If you just tripled your return on your initial BFL product by being an early adopter wouldn't it make sense to buy MORE while the difficulty is low so that when things stabilize you'll have a bigger slice of the pie?

I think the graph of total hash power will look surprisingly similar to the above. Spike, small dip, flatten out with easy rise.

During Jan 2010 to Jul 2011, GPU became more and more popular to mining, their efficiency is 40x higher than CPU, and the availability is not a problem, and the BTC gained popularity, even with such magnitudes efficiency and popularity increase, it still only generate 10x difficulty increase in a 6 months time frame. I think ASIC will not do better than that

I'm not sure if you know which is the "former" and which is the "latter"

frequently. i've seen estimations of the difficulty rise from asic tasking this into account. i belive it was in my thread, "difficulty post asic?" where there was some very detailed math done. should be just a bit down below in this section.

yes however, that wont instantly happen. If the ASIC devices come out full force, there will be a rather long~ transitioning period. Small miners will realize that they need to buy up ASIC devices, but will continue to mine in the mean time anyway until they can trade off their old hardware. Large miners with GPU farms producing Gh/s with their hardware won't be leaving anywhere near as fast.

Has any one mentioned that GPU mining will probably become obsolete and therefore an eventually reduction in the total hashing power as GPU mines are re-purposed?

Good points, I realy dont think it's a "how high is the network going" concern, it's more of a "how quickly will it raise" concern, As, Lets not forget lot's of people want to be on the first shipment, As thats the most profitable time, The slower it goes up, The longer time between block difficulties = more profitable mining.

I also see people selling their early order point, Given a choice of getting a premium for an item, or run it for undetermind rewards, I personally would go for the later.

There is likely many less orders than you would expect from the order numbers. You get assigned an order number when you submit the order, regardless of whether it gets paid for. I know I made one to test that I never sent the BTC for, and it wouldn't surprise me if many people placed orders and then either didn't pay, or waned to change quantities and made a new order.

If so, and assuming a more modest 100 Gh/order. 450 - 500 Th will have been sold before ship date to the tune of $14 million for BFL. Assuming 25% of that is FPGA trade-in, that's $10.5 million in new prepaid sales. I'm not denying it could happen, but it's hard to believe. 

This is slightly lower to the early guesses of 5 single SC per order @ 200GHs, as I think more honest people are saying they can only affored 1-2 Jalapeno's, I realy think it's more like 2-5 to 3 single SC per order, But I think we will see orders reaching 4500-5000 before ship date, So it will all be the same in the end.

I'm hoping that 13Th/70 orders is not a representative average across all 1400 orders, but it could be. It's hard to believe though - this is 186 Gh/order, meaning that people are on average ordering the equivalent of 4.5 SC singles (for about $6k). This would mean that BFL is up to $8 million in sales (including FPGA since those are traded in). This too is hard to believe.   

I noticed by the way there about 70 orders tracked right now.
These 70 orders stand for 12 TH.
Though there 1400 orders placed by now.
What if you take these 70 orders as an average:

1400 / 70 =  20

Added computing power based on 1400 orders: 20 * 13TH =  260 TH

Difficulty increase factor: 260 TH / 12 TH = 21~

New diffuculty: 21 * 1751454 = 36780534

Could this be in any way legit?

Yes Grouver, I remember past comments like this last time, Though i'm more inclined to beleive them this time, As It's been said before, more staff=more units=better ship time.

I really think we have all speculated this one far too much, it's just a waiting game now.

My only reserve (I'm sure you all agree) is why BFL so cagey with figures and dates, I know that cant say for certain, But there isn't even a picture of them? nor power usage or even if it will still use same software as the SHA256? Do we know even what pools its compatable with?

Im not sure if thats my order number, considering I haven't confirmed the order completely yet. That was just automated, however it says my "order has been placed". I'll update when I get another email.

And yeah I read that too.

wow 3063! also i found the lowest order no, garr255 - 1656 <----am I allowed to post name?

https://bitcointalksearch.org/topic/m.1010819

I was also thinking that HOWEVER there is a delay on the older Single Models still. They have 2 month delays to them, so if they don't actually end up delayed the SC Single models, I would only get me Older Single about a few weeks before the SC Single comes out rendering that plan almost useless.
That's assuming they don't delay of course.

Also my email:
Your order from BF Labs Inc. - 3063

Just a view on order numbers, Being intrested in this (not orderd any (yet) but watching its progress) I have seen posts claiming IIRC 1673 as the first order number (on this forum anyway) for a pre-order ASIC, And retrospectivly the last one I heard of was in 2500's area, I beleive its quite possible that some orders are actually for SHA256 as if we assume that there will be a delay in shipping, would'nt it be better to get a SHA256 wait for that, mine on it for a few months, then trade in for a 'sc'

Would it not hypothetically be more profitable to invest the $1,300 for the purchase of the "Single" into BTC itself, and then sell the BTC when these machines come out? Assuming the market responds positively which it may/may not.
The only reason I raise this question is because I'm torn between buying a bunch of Jalapenos (or even a Single outright) or just investing the money into BTC and rolling the dice to find out for instant profit.

Because these SC devices can come out and really make the difficulty level go bonkers to the point where the devices cannot make up profit. Also judging by how BFL is being super hush hush and not actually giving customers much insight to when exactly they will ship, if I were to order now it is quite possible I won't get it by January-February and everything will have adjusted by then.

 It's a very hard situation because, as crazy as it is for someone like me with the budget I have, investing the $1.3k into BTC for a possible 3 month instant profit seems to be a much better idea due to the unknown possibility of how long a machine will get to my doorstep by the time the difficulty increases.

EDIT:
I may just go ahead and buy a single. I'm just really worried that the current "estimation" of ~200-250 days for ROI is incorrect. I guess it goes with the territory of investment.

Whatever difficulty is required to ensure a ROI of +8 months at the current market rate.

If figures are as claimed, if they are able to ship within times claimed and if their units are reliable Id expect the ROI on the new units to shoot past 8 months right out to 2 years and beyond.
After 8 months of ROI they stop being such an attractive investment and orders will subside and BFL will slash their prices to maintain sales volume.

BFL would of made their ROI on the initial round of pre-orders (or close to) leaving them to sell their ASIC design cheaper and cheaper until they would be selling remaining stock at cost. Pricing so that it always  remains at approx 8 months ROI (dropping to a lower ROI if another competitor appears).

It will have very little effect on the market but I think for many miners (especially BFL suckers on pre order) you will see the chances of making a ROI vanish before you even receive your order.

ROI will grow out even further if BFL holds back all pre-orders for 3+ months to pre-mine on them.

If you check the thread about ASIC order numbers, you could get a rough idea of the spread of orders. I know for a fact that there is one confirmed order for 4 SC Rigs already.

The numbers I've been using are generous assumptions towards the worse cases. Example, assuming they could indefinitely ship 50 Singles/day and 3 Rigs every day..indefinitely... quite generous for worse case I think.

Another approach was looking at how many FPGA Singles were probably sold already. I think knowing how many FPGA Singles were sold would be indicative of the core person interested in buying anything from BFL. From the difficulty change from March 2012 around when I'm guessing BFL was getting shipments out there finally, to today, we have only about 2.5TH increase in difficulty. Assume people that were putting their new FPGA Singles online took their GPU's oiffline to accomodate their FPGA's (why they would do that? ...I don't know...but the BFL person ont he forums told me that is what they heard was happening from customers)..and you get 5TH from FPGA Singles. Which comes out to about 3000 to 6000 Singles sold and now mining online. Seems very high, but that is what making a generous estiamtion does. So, that's a worse case for how many FPGA Singles will be traded in. That would require somethign like 3.x million dollars to be sent to BFL for tradeins. The numbers seen going through bit-pay, as I have heard, don't resonate well with that number, imo. Unless people are lagging or particularily need to use bank wire.

One speculative reason to doubt those numbers is because BFL is only up to urder number 2300 or so now. And I think they started around #1600 when pre-orders started for SC hardware. And if that 1600 are sequerntial order numbers since inception of order taking process, then that would mean 1600 orders accounts for 3000 to 6000 Singles shipped? I don't know..maybe... any thoughts on add or about that?

||bit

Well, we know Bitpay brought in $250k the first day. Given the paranoia about being first in line it's likely that most people probably got in ASAP, but let's say there was another $250k in bank wires, and the rest of the preorders after the first day are worth another $500k. That's a cool million dollars.

Given BFL's current price per TH/s (~$32,000), that million dollars would buy 32TH/s.

If you're estimating 200TH/s, BFL will either need to slash prices, or pull in $6m.

I think your worst case is the best case BFL will want as many of these things out the door as possible before people catch on to how badly they are going to be screwed.

I used the observable changes in past difficulty level, not BFL's current shipping capacity to determine the numbers.

Using just the shipping capacity alone, it wouldn't be bad for the sc-rigs coming online at the fourth week. Take the worse case of 3 x 1TH rigs/day. That is 18 or 21 TH/week (I doubt they run 7 day weeks though). Round it to 20TH/week to approximate a worse case. After four weeks, that is 80TH more online. With that increase, a new rig just coming online would still pay for itself in about 4 months because the difficulty would only increase by a factor between 6x and 7x. BUT, of course, I think you'd want to put it in the context of ALL the other orders that would eventually increase network hash rate. So, how large are those orders? Not sure.... So, let's just make it easy and use current shipping capacity as we have been told hey are filling now. They are shipping 50 Singles per day and 1.5rigs per day. Assume the same rate for new SC Singles, and let's increase the rigs to 2.5/day for a more worse case in the below scenario. And assume it will somehow sustain that rate indefinitely.

Assume 50 Singles/day and 2.5Rigs/day from the start of shipments. Exclude any built up supply for now.

50 Singes = 50 x 40GH = 2TH/day shipped.
2 Rigs = 2 x 1TH = 2TH/day shipped.
Total: 4TH/day or 28TH/week or 112TH/month

After four weeks that is 112TH + 12.5TH = 124.5TH. Two months would be 124.5TH + 112TH or 236.5TH on the network. Carrying it out like this out... assuming they somehow kept such consistent orders... Some math I did suggest you'd still pay off the rig (unlike Zeno's paradox but not that fast). The calculation yielded about 14 months to break even.    And again, not counting the btc discoveries going from 50 to 25...so, maybe [much] longer if the btc value doesn't compensate for that factor.

||bit

You leave out the th/s boxes again one a day is seven th/s a week added, two a day fourteen th/s a week, three a day twenty-one th/s a week added and so on.... The first to get these things should be able to pay them off and make profit, third or fourth week receivers of them good luck, second and third month not a hope in hell of seeing your money back.

That's true. However, if we just use the dollars we think are interested in SC hardware, we can probably make an fair estimation regardless of what specific models they ship. The hashing power per dollar is close enough between the tiny 3.5GH Jalapeno and the 1TH SC mini-rig. So, to kinda average it out, I tend to just focus on the middle valued product (SC Singles).

Jalapeno:  ~$43/GH
SC-Single: ~$32/GH
SC-Rig:     ~$30/GH

||bit

Can you give your objective basis or reasoning for your estimation?

Consider the difficulty level now and over the time that BFL has been shipping FPGA Singles. It has only increased ~2.5TH since about March 2012. Assuming worse case that is all increase due to Singles, and generously that each Single replaced a GPU that went offline, that would be maximum about ~5TH increase of partial hashing power due to the FPGA Singles.

Each FPGA Single is 830MH/s. Each new SC Single is 40GH/s.  That is a factor of approx. 48x increase in hash rate. Scale up the presumed 5TH of partial hash power of FPGA Singles by 48x and you have 240TH worse case scenario. So, 120TH to 240TH increase from all tradeins (assuming all are traded in for SC Singles). Count new straight orders..hmm you might be right. :/

Currently 12.5TH increase by 120TH or 240TH:
120TH + 12.5TH = 132.5 TH.  Difficulty increase 132.5TH/12.5TH = 10.6x
240TH + 12.5TH = 252.5 TH.  Difficulty increase 252.5TH/12.5TH = 20.2x

So, we are talking a x11 to x20 increase in difficulty in possibly the first month or two of shipments. More than I thought.

Even so, Each type of SC hardware is about the same price to hash rate, so it doesn't matter much on what you buy. And at 20x difficulty breakeven may reasonably be expected around 4 to 5 months of continuous mining. And probably much faster for those that recieve shipments first. Don't get me wrong, 4-5 months is a good ROI. But I didn't count the BTC blocks from 50 to 25. If the BTC value goes up as a result then no problem. If not, your talking upwards of about 9 months ROI (still not bad). And finally, if new orders (i.e. non-upgraded tradein orders) are the quantities as trade-in/upgrade orders. Then we move to an approximate 20x to 40x increase in difficulty. ROI would then be either about 9 months or 18 months.

||bit

I estimate that once all BFL FPGAs are exchanged for BFL ASICs, the processing power will be atleast 300TH. Also reward halving will make it 600TH by today's standard.

At this point the 40GH $1299 unit does not look so cheap anymore, as it will take around a year to pay itself off.

Ofcourse if the price per Bitcoin rises then it will make things easier.

BFL said that were going to ship all three products at once so count on some of them th/s boxes going out the door ever week if they ship any amount of them a day then the diff is going to ramp up real quick.

For those interested, I'd like to collectively see if we can determine some reasonable profile on how BFL asic may come online and how it alone may affect the difficulty from the initial shipments through subsequent months. I'll list some data and assumptions that may be helpful. Feel free to critique them and provide any other ideas or  observations that you think might be helpful.

As I understand it, difficulty is directly proportional to network hash rate (hereafter:NHR or nhr). So, if the nhr increases 120% then difficulty increases 120%.
Current nhr is ~12.5 THash as found here: http://bitcoin.sipa.be/

Regarding recent pre-orders:
Does anyone have an idea on what the first order number was in the recent pre-orders or some order number that is probably close to it? I've read some numbers around #1600. However a recent order number I saw was around #2300. I have a hypothesis that the number 1600 is just picking up from where the fpga Singles & fpga Mini-Rigs have been ordered. If we could get an average order size, then we could figure out how many Singles have been sold. And then reason that this is how many SC Singles will be ordered by tradeins alone.

One observer in a thread in the forums pointed out how the Bit-Pay (the initial pre-order method of payment) spiked in the first 24 hours. If I am not mistaken, that spike was about 50k-ish bitcoins. Using that value over a projected ten days, I did some rough math then and it came out about 8TH of BFL Sc hardware if there would be ten such days of products beign ordered(assuming the SC Single as the average H/$ value - i.e. 40GHash/$1299.

Other stuff:
* BFL has been late in shipping products. Shipping products on average about 8 weeks (can anyone verify this?)
* BFL has about three months to get some initial inventory built and ready to ship in October.
* BFL claims to have shipped 50 units per day. However, looking at past difficulty changes over the montsh does not indicate this. One BFL rep on the forum said this was b/c people were simply in large replaceing GPU's with their Singles. However, this does not make since for a miner to do, since GPU's are still arguably the fastest mining hardware. It seems more likely that this is a current shipping rate as also indicated by one BFL comment. So, we can assume a current shipping rate of 50/day (Any other thoughts on this?)
* BFL is ramping up staff (witnessed by Inaba). Also, BFL has in emaisl and the forum indicated that they were ramping up shipping. So, this may more solidify, if not improve, the above 50/units per day.

Any critiques or thoughts? Or it may be as simple as using the current production and shipping numbers and assume that only SC Singles are shipped, with a initial surge of one or two months product production coming online within the first couple weeks initial shipments are sent out.

||bit