/diy/ - Do It Yourself

A link to the mining rig website

https://products.butterflylabs.com/homepage/4-5gh-bitcoin-miner.html

It's impossible to make money. They best you can do is break even on your electricity bill. There's no use to this unless you're planning to buy kiddy porn or drugs with it.

>>418501
if you are using a butterfly labs product it uses next to nothing on electricity. i called BFL and asked how long the back order would be on a mining rig using ASIC chips. they told me i would revive it anytime between may-june. the back order it just too long. now that asic based mining rigs are online the price of bitcoins will be dropping (or that's what i believe). its spending your money and not seeing a return for months because they are on back order.

TL:DR- no i don't think its profitable because of the unknown.

>>418508
>>418508
Right but even then the price of them right now means that i would only need to farm four coins to make up for this product. Even if the price drops to $20 each i would think that i could recoup my investment in a month or two after i received this mining rig. Even if the market lowers in value by then its still a near garentee that i make my money back at the very least?

>>418538
If you can predict the future there's better ways to make money than buttcoins.

>>418539
Right but can you deny the fact that this device will pay for itself given the time? No matter how long it takes it still farms coins that are redeemable for cash so therefore it is a garenteed investment

it's sad but funny to see people who think this sort of speculative investment is a sure thing, but are not able to see that other speculative investments can provide a better return.

people like OP are like those who play the lottery, which is gambling for people who cannot do math.

>>418541
It may do so, but it is by no means a certainty.

Im starting to see bitcoins represented in tv even

>>418543
Look im just looking for cashflow wherever i can legally, this is a work less investment that can either make huge returns or tiny returns, from what ive read the bitcoin market has taken huge hits before and still survived. The coins may lose all value even but if the market picks back up then it can still be a good investment. I have other ideas for income other than a 9-5 but those require their own investments, work to physically assemble and create, dev costs, and then i have to find the market for them. This is a one time investment that could potentially pay for all of that in time or i could be out $150. Given that $150 and really be invested for much else that would give any returns (unless i sell drugs) then i could either spend it on food and gas, or on hope

>>418550

ok, chief, i'll try once again.

the only miners who made money got in early. just like the people who bought apple stock 20 years ago. they are all rich now.

your only options are to find the new bitcoins, or the new apple. get it?

in other words, get a damn job like the rest of us. you are chasing taillights.

god damn I wish I had bought into this shit when BTC were still at like 70 cents.

>>418560
I get what you are saying but i dont see why mining them isint a viable option

Im not looking to get rich, im just looking to make some sort of imcome at all from it

>>418586
Professional miners already have gobs and gobs of GPUs and FPGAs and ASICs humming away. You're small time and are unlikely to find a new bitcoin on your own. Your best bet is to join a pool and you'll get peanuts.

Remember that when there's a gold rush on, sell shovels.

>>418592
Ok, thats an analogy that i can understand. I already know that pool owners arent obligated to actually distribute any of the returns

>>418560
>>418592

Very well said, both of you.

I personally know a guy who has made a good amount of money off bitcoins, and he did so by using his employers systems at a data center he worked for (he had a management position so was able to get away with a lot of prohibited shit.) He had hundreds of gpus running, all the hardware and electricity was being stolen from his employer and he pocketed 100% of the bitcoin profits.

There's definitely an allure to the idea of making money off of bitcoins, I mean I remember a year or so ago I was tempted to buy a few hundred dollars worth when they were selling for ~$9/each and now, as you said, they're trading on mtgox for ~$45. Nothing compared to their initial release when they were worth something like $0.005/coin or some shit - honestly a $20 investment back then would be worth hundreds of thousands to millions now.

>>418624
Exact reason why i dont see a huge harm in getting a small system to passivly mine for me, if i can even mine out 10 coins then i would be happy, the value only seems to rise and mtgox shows that they are reletivly stable for the most part. It could be nice to just have a few going for a rainy day, or for the far future when they are possibly worth over $100 each. Plus if the market takes a huge hit i would be down to purchase as many coins as i can on the chance that the market will pick back up again

I'm an EE, i was iniitally thinking of buying 2 or 3 singles (~$2600-$4000 worth) of ASICs

However I have absolutely zero confidence that BFL is capable of making those boards. Everything we have seen from them says that, if they work at all, they won't last very long. Their PCB design sucks (through holes on every pin) and their FCBGA have bad epoxy filling, meaning bad thermal conduction to the boards.

>>418684
Ok so a surface mount board would be best for these? I have some experience in PCB design so treat me like a pro, that way i can feel stupid and learn.

Bit miners from what ive read alone in this thread require a LOT of electricity no?

Also would it be possible to design my own PCB specifically for mining purposes given the software and a schematic diagram of the circuit that needs to be made?

>>418691
It's a reflow assembly and they have non-filled, through hole vias under each and every pad of their flip chip ball grid array asic

they run the risk of the solder connection between the board and the IC breaking, as solder could reflow into the via hole, leaving not enough between the top layer pad and the chip. especially when combined with thermal cyclical stress / expansion.

better would be to A. not use vias on every pad or if you have to B. coppper fill the vias so the solder paste sits on a flat plane

BFL said they had a 10 layer board. I don't doubt it. It would be very difficult to make it in less than 6. It would probably be next to impossible to meet FCC EMI/EMC specs with less than 8. (Though i guarantee they did not even try to meet or test that.)

they're already in violating of FTC laws about preorder promises, why not violate FCC laws too? the liberatarian paradise known as bitcoin doesn't care for "big government" (until they get scammed and run to consumer protection laws)

additionally now BFL just switched to "all orders final!!" and i don't think they've even managed to get the system hashing yet (they CLAIM it's due to firmware bugs and not the chips themselves.)

Even if they work and you get them, i have serious doubts they will last even a year. And their "warranty" has no real backing. They'll probably fold up and run away with whatever money they got once shit starts to hit the fan.

>>418697
Ok i actually understood the issues with the board. no way i could make a 10 layer though seeing as how my class only did a single layer. I might however know someone that could do it (a professor at my local community college that used to be a lead circuit designer for intel)

But i do wonder how easy it could or could not be to break into the shovel market so to speak. Are there any competitors to BFL or are they basically a monopoly in the shovel business?

>>418691
>>418697
ASICs are very efficient, they actually only draw a few watts per chip. BFL claims i think 60 GHash/sec for 60W.

[they of course outsourced the 65nm ASIC design, they're nowhere near being capable enough to design such a thing]

designing the PCB would be difficult enough as it is with so many high pinout low pitch components and all the signals needed to be routed (probably many of them controlled impedence to some extent) but there's no "schematics" for this lying around to my knowledge. The ASICs are basically just SHA256 engines, you still have to put the other stuff on the board to control everything, and i'm not as familar with that.

>>418701
it's really not worth it. the hard part isn't laying out the PCB

the hard part is getting an ASIC designed that sufficiently beats competitors and will have enough buyers to justify the several milliion dollar NRE (non recurring engineering) cost of designing such a chip

BFL outsourced a 65nm and got iirc 10Ghash/watt

you'd basically have to go to at least a 45nm and probably aim for at least 15 GHash/watt in large quantities to be economically feasible.

there is someone on bitcointalk who, thus far, appears to be legit and works with engineers making a 28nm chip, capable of efficiency close to 10x higher than BFLs. He is not selling it to people, but looking for companies to sell the chip to.

I'm waiting to see what happens there.

There's not that much room for profit in this area though imo. one time costs are very high.

>>418701
>>418722
avalon asics started way after BFL (bfl originally took preorders in like july of last year?) and they have shipped two orders, totalling i think about 300-400 ASICs, to people. they were the first to customers

ASICminer is another group that made their own 110nm ASICs but they didn't sell to people. they sold shares in their mining, and they've been doing it for about a month or two now.

>>418699
Isn't it a major point of the buttcoin design that the computational complexity increases at roughly the same rate as the computational capacity of the network, thus in essence balancing itself? If ASICs become popular, coins become more expensive to mine, and unless you were an early adopter you've gained nothing.

>>418723
>avalon asics
Well i already stupidly typed out a business proposition to the professor and sent it. Any chance that a lead chip designer for intel could design somthing workable for profit using existing hardware on the cheap?

>>418727
the sum of all the profits in bitcoins are pennies to intel. they don't care, and they shouldn't.

Their investment into their world class fab is on the order of $10,000,000,000

But if they wanted to, yeah, they could probably get ~20-50x the efficiency of the 65nm BFL chips, even at 28nm.

https://bitcointalk.org/index.php?PHPSESSID=dtmbmbbk7rso6iecm67i17tck6&topic=146371.0

this is the thread i'm keeping an eye on. If that guy is legit, and not just a scammer / pulling everyone's leg to manipulate the market, they are going to be the chips to beat.

of course having ANY asic at all will probably still be profitable for a year or maybe even two. but as more asics and more hashing power comes out, you start to really care about asic vs asic power efficiency.

as is basically any asic is 50x or more efficent than an FPGA and probably thousands of times more efficient than a gpu.

>>418729

I said used to be, these days my guy is a professor at a community college in his late 60's to 60's and plays in a band while helping manage the robotics club, a generally awesome guy. My question was not about intel themselves but if someone with his experience and skillset could design something that could compete with the rig making companies.

My issue here would be convincing him that i deserve to be paid at all given that he could do it. Seeing as how im nothing more than a student in the program that has yet to complete it

>>418729

>implying ASIC are easy to produce correctly
>implying a FAST FPGA is cheap
>implying a top of the line GPU for scientific processing isn't THOUSANDS cheaper than either of the two above and is more supported.

>>418746
i have no idea what you're trying to say because i agree with #1, for #2 i made no mention of FPGA except to correctly point out they are far less efficient than asics, and #3 .. if you're talking about using a GPU for bitcoin minings, just go look at hash rates between GPU vs FPGA vs ASIC and most important, Hashes per watt.

>>418748

You were talking about power efficiency and performance.

I agree that ASICs and FPGAs are fast as hell. But ASICs are expensive to design and produce and a GOOD FAST FPGA is pretty damn expensive as well and hard to get in a small order.

GPUs on the other hand are easy to get, "cheap" in comparison, pretty damn powerful and have a lot of support (Sample code, issues with errors/irregularities etc. etc.).

Hashs per watt though.. I pay like 8.5c per kW/hr... It will take me an eternity to recoupe my losses by switching to an ASIC or FPGA.

If you're a fairly large well funded organization/group.. ASIC is your fastest bet.

If you're a academic group of some sort with a good budget, you can probably get your hands on a top of the line FPGA.

But for normal guys like OP and me, our best bet is a GPU.

>>418753
FPGAs have been mining and being sold for a whjile now. They're old hat at this point

ASICs, when they begin to become widely distributed, will raise the total hashrate of the network by so much it will force GPUs out, followed by FPGAs.

can someone answer my question about the possibility of having the professor at my college help set up a small mining rig business? Would an ex lead circuit designer be capable of innovating something efficent enough to produce a decent to good bitcoin mining rig?

>>418753
ASICs are expensive to design, but the per-unit production costs are low.

>>418754

I'll just leave this here... (It is an older study, but the demand for GPU scaling is much higher than FPGA I'm sure.. More market.)

http://cas.ee.ic.ac.uk/people/ccb98/papers/DavidFPL10.pdf

FPGA vs GPU High Performance Computing:

FPGA:
"The HC-1 is a 2U server card that uses four Virtex-
5’s as application engines (AE) to execute the distributed
processes. The HC-1 also uses another Virtex-5 for process
management and eight Stratix-II’s for memory interfaces.
Figure 1 shows their arrangement.
The resulting system has 128GB of DDR2 RAM with
a maximum bandwidth of 80GB/sec."

GPU:

"The GTX285 is a member for the GeForce 200b family of
GPUs made by Nvidia. It has 240 core processors running
at 1.4GHz and supports up to 4GB of external DDR3 RAM
(we use a 1GB version) with a maximum bandwidth of
159GB/sec. Each core is grouped with seven others and
two small but low latency memories: shared memory and a
texture cache. In contrast to the shared memory, the texture
cache is optimised for 2D spatial locality and can access
the global memory directly."


>pic related it's the results

"The HC-1 is at a significant disadvantage in terms of
cost, power and floating-point operations per second. The
GTX285 is a significantly less expensive, uses half the power
during operation and can theoretically perform at 1062
GFlop/s. However the GPU has a fixed, 32-bit architecture
and the only way to synchronise all of its threads is by
stopping and restarting its kernel externally."


"For most
of these the GPU significantly outperformed the FPGA
architecture. The one exception, the generation of pseudorandom
numbers, used closed-source firmware customised
for both the task and the platform. We suggest that without
a standardised FPGA HPCS platform about which opensource
firmware could be developed, the future for FPGAbased
HPCS will be increasingly marginalised to specialist
applications."

>>418759

And therefore you'll edge out in the long run, IF you mass produce and sell a shitload of them. Otherwise you spend a large chunk of time/money on a one off investment that you hope works correctly.

>>418760
Too bad the paper didn't test the kind of computations relevant for buttcoins.

>>418760
I'm not talking about hashing power, i'm talking about hashing efficiency.

That is the limit.

That's what made CPU mining economically unfeasible, and the coming ASIC swarm will make GPU mining unprofitable as well

>>418761
You want to be the shovel-seller.

>>418756
The problem is not skills, really. The problem is money, i.e. paying for a 45nm mask set or share of one.

>>418768
it's not trivial to design said 45nm mask set either.

>>418764

>SHA-2 Hashing
>Split message into 512-bit blocks
>Put in array
>Take 4 byte chunks to make 32 bit integers
>Perform operations on said 32 bit chunks (in an array)

"Too bad study didn't look at 32 bit operations."

You went full Potato.

>>418746
>fast FPGA vs BIG FPGA
>has no idea how FPGA's or hardware synthesis works

>>418756
No, because circuit design is about 1% of the challenge. It's more about programming the FPGA(s) and a thorough knowledge of HDL and simulation.

>ASICs are expensive to design
Designing isn't necessarily expensive, and certainly not the most expensive part. Anyone with software who knows what they're doing can pull a viable ASIC design for a specific purpose essentially out of their ass.

Pretty much unless you're an engineer who has academic background in the subject, you probably don't understand most of the concepts or issues surrounding FPGA or ASIC performance.

For example how many here have actually programmed an FPGA?

>>418775
the whole point is irrelevant because the true arms race is hashes per joule. asics are incredibly high hash rate and incredibly efficient, so they are going to raise difficulty and total hash rate by so much they may even force FPGA miners out entirely.

>>418771
No, I think it's pretty straightforward since you'd just use standard cells. Software licenses cost a fair bit though.

>>418778
I've done some HDL in school but i didn't particularly care for it and don't work with it in my field (RF / Medical)

I think the limiting factor for the ASIC design may be getting good yields at an efficient node, really more of a process issue than a complicated design.

>>418775
The most relevant test in the paper is the random-number generator, and it references another paper showing a better FPGA implementation outperforming the GPU by a factor of 8.

>>418782
It's not trivial for these jokers and frauds like butterfly labs, let me put it that way.

If you've been doing VLSI for a few years and have access to the tools, resources and knowledge/experience capable of doing it, yeah, then your biggest issue is probably getting the money for the NRE and maybe contracting someone to do the PCB Assy (or just sell the chips yourself like that guy on bitcointalk seems to be speaking of,.)

but these people aren't EE's.

>>418760
But in that application, the FPGAs are essentially soft processors...... it's not comparing what you seem to be thinking

>>418784
>but these people aren't EE's

Heh, yeah, I guess that's probably the crux of the matter.

>>418783

>implying you hash with a random number generator

I usually hash sequentially.. starting from all possible combinations of x values up to all possible combos of y values.

The actual hash itself doesn't give a fuck about random numbers. The hash function has nothing to do with random number generation.

>>418787

What do you think is going on here?
What do you think hashing is?

>>418789
Oh wow, talking about potatoes. Both algorithms are centered around bit-twiddling. Reading comprehension doesn't seem to be your strong suit?

>>418792

Yeah.. and this has what to do with random number generation?

>>418794
Reading comprehension REALLY seems to be an issue for you.

>>418771
So i basically just played the fool on that one by sending the prof an email? lol awesome

>>418791
all those tests are done using a processor architecture implemented on the FPGA's. Not something more efficient like a large parallel multiplier implemented as hardware. You essentially end up with pin-pin delay as a limit in FPGA speed if you have enough LE's and space to route connections well.

Do you understand? This is pretty elementary stuff.........

bitcoin mining works on the fact that sha256 (and all well designed hashes) are extremely sensitive to the input so flipping one bit leads to a massive change in the hash output (due to avalanche iirc)

so you cannot effectively map input to output at all, meaning that no matter what you hash it's basically a random output (i.e. can't predict it)

therefore the payout is set by whoever randomly gets lucky enough to get X amount of 0';s in a row on the hash output (something like 50 now i thought?). the only way to do this is brute force hashes.

that's my understanding, obviously if everyone started hashing from 1 it wouldn't work, so they probably have some method for doing it, i think hashing is how everything in the network gets "approved" or whatever.

>>418796

Dude, you seriously have no fucking clue what you're talking about...

The hash function has NOTHING to do with random number generation. It is shifting and rotating of the 32-bit chunks and shoving them back together.

>>418799
What's LE again? Logical something?

>>418800

When you generate a random input, how do you know you aren't accidently trying to same input over and over?

Keep a record somewhere? Put it in a collection in the program?

"OUT OF MEMORY EXCEPTION"
"THE DESTINATION DEVICE IS FULL"

>>418801
Have you looked at the Mersenne Twister and SHA256 algorithms? Anyway, my point was that of the benchmarks used in that paper, that one is the most relevant for buttcoins, and coincidentally the GPU and FPGA performance was equal.

>>418808

Neither of those algorithms use random number generation, hell the first one IS a pseudo-random number generator.

>>418808
as pointed out before, that paper is almost 100% useless in drawing comparisons between devices in the specific application of bitcoin mining, and in fact completely removes the reason that FPGA's do so well in said application (i.e. optomised custom hardware description vs. generic processor architecture)

>>418810
What the fuck are you on about?

>>418811
Which was one of my points.

>>418812

What the fuck are all these people talking about random number generation having to do with hash functions on about?

>>418814
Both involve a lot of shifts/rotates and bit operations. Therefore the expected performance of the two loosely correlate.

>>418818
only pRNGs. True RNG has no bit operations

YFW GPU'S, FPGA'S = ASIC'S

where is your god now?

>>418822
That was a 100% pointless and irrelevant comment. Well done.

>>418824
No it wasn't. There is an important distinction between them. They are called PSUEDO random number generators - pRNG for a reason. As opposed to (true) random number generators - RNG.

>>418825
Still doesn't make it relevant to the discussion.

>>418821
:DDDDDD

I can hash faster using high speed cmos logic.

All with discrete components...

>inb4 I'll out compute you with diodes/transistors/resistors

>>418883
>I can hash faster using high speed cmos logic.
>All with discrete components...

b-b-but the pin-pin delays will end up being massive since you have to make sequential logic with them. Also if you're using 74HC discretes it'll end up being an asynchronous piece of shit due to lack of clocking. Similarly, the transistor idea doesn't work due to the high gate charge of discrete transistor components meaning much higher delays than ones implemented in IC's. It gets worse going to even the fastest diodes for diode logic as well. Also imaging the power consumption and PCB layout issues you'd have when you actually try to put it together.........even with optimal routing you're traces would be absurdly long and high inductance

>>418959

>implying the pins aren't wired straight too each other in a convoluted ball shape
>implying the entire ball isn't cryogenically froze
>implying there isn't a grounded shield between every pin to reduce interference.

Just cause you don't know how doesn't mean you can throw me off my project.

If this doesn't work, I'll just use an Arduino.

>>418962
>cryogenically froze
b-b-but semiconductors don't work like ohmic resistors....... lower temps typically increase the activation energy to allow majority charge carriers in p-n junctions to move hence IC's only work over a certain temp range and are generally optimal at near room temps. Also based on the sheer number of chips you'd need to synthesise the logic present in hashing algorithms, the the mass of your 'convoluted ball' would result in the requirement of copious highly intricate structural members being present throughout the design. Even with the grounded shields, you'd probably want to use buffers to level shift the CMOS logic signals to reduce noise to a level where glitches are infrequent

>>419031

Oh you poor poor child. You assume I'm using your foolish silicon based Semiconductors?

http://en.wikipedia.org/wiki/Rapid_single_flux_quantum

Your 5-10 ns second delays are an eternity to me.

Welcome to 5 picoseconds.

>>419038
>http://en.wikipedia.org/wiki/Rapid_single_flux_quantum

so I hear development of this idea stopped like 15 years ago? Was there ever a functional device made using this concept?

>>419041

See the bottom of that page

Superconducting Technology Assessment, study of RSFQ for computing applications, by the NSA (2005).

NSA gets a good idea 2005. Project drops off the map.

What do you think?

>>419042
I can't find an example of a single functional device or even junction having been constructed. This seemed to be all the rave around 1995 going by the number of academic papers that pop up in my searches. Some of the delay figures quoted in papers from the 90's are achievable by modern RF and microwave conventional semiconductor components. I think there may have been some issue on the theoretical level rather than manufacturing limitations preventing the device being made.......though I haven't read much

>>419045
It's just Josephson junctions, so nothing really new. The problem is scaling it up to thousands of junctions so you can actually do something useful with the logic. At that point it's probably cheaper to go with GaAs, like you point out.

>>419045

I don't know about RSFQ in particular, but there was some FFT processor chip based on Josephson junctions years ago. It was quite fast, but apparently interfacing to it was a major hassle.

>>418624
Do you even into math?
>$20/ $0.005 (per BC) = 4000 BC
>4000 BC * $45 (per BC) = $18,000

>>419031
>not using qubits

>>419215
180,000

jackass

>>418495

>implying Bitcoins are easy enough to crack that you'll get anything at all.

Not with the amount of stupid faggots out there doing it now.

>>418495

Bitcoin mining is like pissing in the ocean when you have a prostate problem.

oh god I wish I had bought some when they crashed a while ago...

>>419420
Don't worry, it'll happen again.