Official Open Source FPGA Bitcoin Miner (Last Update: April 14th, 2013)

[Jaime Frontero]

Nice, and many thanks for the release.
Now, some rough calculations.

At 80 MHps, I will need at least 3 of these to achieve a single 5830 hashrate.
That is $595.-x 3 = $1785.- at full price, vs. $190.- for the 5830.

Giving the 5830 is consuming $11.- a month in electricity, and assuming this board will consume zero electricity, it will take more than 145 months, or 12 years to recover the investment, always comparing to a 5830.

Is it really cost feasible ?

Your math assumes the difficult will be constant.

Eventually, the difficulty will reach a point where the power cost is a significant portion of the mining cost.  Right now, power is about 10%-20% of the cost when doing GPU bitmining.

Let's assume that one year from now, power is 80% of the mining cost.  At that point, FPGA mining would have a significant advantage over GPU mining even if the hash rates are same.

my power cost right now is about 4-5% of what i mine in Bitcoin, at $7/1BTC.  1 - 1 1/2 days of mining pays my monthly electric bill (that 80% of it which is incurred by mining).  my rigs are paid for.  it's not as grim - and won't be - as you make it out, IMHO.

[1713929395]

1713929395

[1713929395]

1713929395

[1713929395]

1713929395

[1713929395]

1713929395

[eturnerx]

That's fpgaminer for opensourcing the code. Imma start hacking away!

[gusti]

Nice, and many thanks for the release.
Now, some rough calculations.

At 80 MHps, I will need at least 3 of these to achieve a single 5830 hashrate.
That is $595.-x 3 = $1785.- at full price, vs. $190.- for the 5830.

Giving the 5830 is consuming $11.- a month in electricity, and assuming this board will consume zero electricity, it will take more than 145 months, or 12 years to recover the investment, always comparing to a 5830.

Is it really cost feasible ?

Your math assumes the difficult will be constant.

Eventually, the difficulty will reach a point where the power cost is a significant portion of the mining cost.  Right now, power is about 10%-20% of the cost when doing GPU bitmining.

Let's assume that one year from now, power is 80% of the mining cost.  At that point, FPGA mining would have a significant advantage over GPU mining even if the hash rates are same.

Sorry, I do not see your point that power over capital investment cost changes with difficult.
Can you please ellaborate or give a practical example ?

[kokjo]

just a optimision, from a noob:
is it possible in sha256_transform.v to remove the generate loop, and replace it with some sort of pipelining instead, it will from what i can understand make the code smaller, but have a greater delay. and with smaller code you could have more of them on a single board, and a higher clockspeed.

am i talking gibberish, or do i have something right?
its was all from what i could learn from the internet in about 5 hours.

i just think it would be an improvement

[greenlander]

Sorry, I do not see your point that power over capital investment cost changes with difficult. Can you please ellaborate or give a practical example ?

Today, you buy $2000 worth of hardware.   Your hash rate is 1500 MH/s.  You draw 1200 watts.  A  bitcoin is worth $7.  Electricity is $0.10/kWh.  You gross $1228 a month and your electricity cost is $86 a month.  Your electricity cost is just noise.

Imagine that a year from now, the difficulty is tenfold.

At this point a FPGA rig is going to have a fraction of your power consumption.  At that point:
Your hash rate is 1500 MH/s.  You draw 1200 watts.  You gross $123 a month and your electricity cost is $86 a month.  Your net profit is $37/month.  Your electricity cost consumes most of your gross.

The guy with the FPGA array has the same hash rate of 1500 MH/s.  He draws 300 watts.  He grosses $123 a month and his electricity cost is $21 a month.  His net profit is $102 a month.

Will difficulty rise tenfold?  Will bitcoins continue to be worth $7?  Will 1500MH/s of hashing power only draw 300 watts in FPGAs?  Who knows?  I'm just pointing out the scenario where the FPGA guy can have a huge advantage.

but at that "reach point" you need FPGA-utilising solution with multiple FPGA's per board or use them into mini-boards[ISA bus ? ], stacked into main one, to minimise communication/power/connectivity impact on solution cost.
and also, board must be ready to put into industry-adopted-sizes racks[both IT and manufacturing/science racks meant]. if you plan to sell number of such boards.

Well, yeah.  I don't think that anyone is proposing that FPGA reference boards are the way to go.  The way to go is to do your own board design and load it up with bucketfuls of FPGAs.

[gusti]

Sorry, I do not see your point that power over capital investment cost changes with difficult. Can you please ellaborate or give a practical example ?

Today, you buy $2000 worth of hardware.   Your hash rate is 1500 MH/s.  You draw 1200 watts.  A  bitcoin is worth $7.  Electricity is $0.10/kWh.  You gross $1228 a month and your electricity cost is $86 a month.  Your electricity cost is just noise.

Imagine that a year from now, the difficulty is tenfold.

At this point a FPGA rig is going to have a fraction of your power consumption.  At that point:
Your hash rate is 1500 MH/s.  You draw 1200 watts.  You gross $123 a month and your electricity cost is $86 a month.  Your net profit is $37/month.  Your electricity cost consumes most of your gross.

The guy with the FPGA array has the same hash rate of 1500 MH/s.  He draws 300 watts.  He grosses $123 a month and his electricity cost is $21 a month.  His net profit is $102 a month.

Will difficulty rise tenfold?  Will bitcoins continue to be worth $7?  Will 1500MH/s of hashing power only draw 300 watts in FPGAs?  Who knows?  I'm just pointing out the scenario where the FPGA guy can have a huge advantage.

In your example, you are overriding the initial FPGA costs. At this moment, for a 1500 Mhps FPGA array, you need 18.7 x 80 Mhps FPGA, or $ 11100.- Same 1500 Mhps is achieved with $2000 worth GPU's. So, assuming a profit gain for FPGA's of 102-37 = 65 a month, you still need almost 12 years to recover initial FPGA investment.

[ktion23]

just a optimision, from a noob:
is it possible in sha256_transform.v to remove the generate loop, and replace it with some sort of pipelining instead, it will from what i can understand make the code smaller, but have a greater delay. and with smaller code you could have more of them on a single board, and a higher clockspeed.

am i talking gibberish, or do i have something right?
its was all from what i could learn from the internet in about 5 hours. Smiley

i just think it would be an improvement

is it posible to "restart" the calculus,

btw is it transmiting all the hash, if it is, wouldnt it be better to transmit a signal if the H was 0?

[fpgaminer]

First off, many, many thanks to the person who made a donation while I was asleep. I'm assuming it was because of this project? Either way, you have my gratitude! It makes me feel that my efforts on this project are appreciated.  

just a optimision, from a noob:

kokjo, thank you for your suggestion! I appreciate all feedback on this project.  

is it possible in sha256_transform.v to remove the generate loop, and replace it with some sort of pipelining instead

That is indeed what the Verilog compiler does with a generate loop. It unfolds the loop, replicating the logic. So that sha256_transform module actually contains 64 instances of sha256_digester.

btw is it transmiting all the hash, if it is, wouldnt it be better to transmit a signal if the H was 0?

Hello ktion23. Thank you for your feedback and interest! It only transmits hashes where H is 0. In fpgaminer_top, this is the code that accomplishes that:

Code:

		is_golden_ticket <= hash2[255:224] == 32'h00000000;
		old_golden_ticket <= is_golden_ticket;
		if(is_golden_ticket && !old_golden_ticket)
		begin
			golden_nonce <= nonce;
		end

Speaking of which, I need to clean that up a little bit. old_golden_ticket isn't necessary at all. I'll go do that now ...

EDIT: On a more personal note, this past week has been pretty rough. Some bad mojo going around the company I work for, which is always stressful. But seeing this interest in the project, feedback, and that donation this morning have made be feel a lot better. So ... simply, thank you!

[ataranlen]

I'm probably going to pick up a Spartan 3E 3 E 250K, so i'll jump on this FPGA bandwagon as well!  You just made getting started alot easier!

[teknohog]

I'm probably going to pick up a Spartan 3E 3 E 250K, so i'll jump on this FPGA bandwagon as well!

A Spartan 3E 250K has 5508 logic cells. If I understand correctly, this is not enough for a design that needs 90K LUTs, though I'm still learning about all of this.

[fpgaminer]

A Spartan 3E 250K has 5508 logic cells. If I understand correctly, this is not enough for a design that needs 90K LUTs, though I'm still learning about all of this.

I'll try to put my serial design online as well. That fits into 3K or 4K I think, although it's obviously much slower (64 cycles per hash). It's a nice toy for people to play with if they have smaller boards.

[kokjo]

A Spartan 3E 250K has 5508 logic cells. If I understand correctly, this is not enough for a design that needs 90K LUTs, though I'm still learning about all of this.

I'll try to put my serial design online as well. That fits into 3K or 4K I think, although it's obviously much slower (64 cycles per hash). It's a nice toy for people to play with if they have smaller boards.

well that would be very nice, so all us with no big money could try too

[greenlander]

well that would be very nice, so all us with no big money could try too

If you just want to fool around, start with this:
http://www.xilinx.com/products/boards-and-kits/AES-S6MB-LX9.htm

It's just $89.

You wouldn't be able to unroll the loop at all, though.  (fpgaminer fully unrolls the loop so it's heavily pipelined.)  You'd probably have re-use the same hardware for every step of the algorithm, so you'd get about 1/128th of the performance.  If you were clever you might be able to do more than that and have around 4 pipeline stages or so, then you'd get 1/32th of the speed.

I'd respect you if you got 1 MH/s out of it.

Still, it could be a fun project for $89.  Just don't expect to do any serious mining.

It appears to come with a crippled version of their tool chain.  However, you can download the full Xilinx tool chain at (ahem) substantially below retail value over bittorrent.

[allinvain]

This breaks my heart, open source FPGA mining code, yet the hardware is prohibitively expensive for the performance that it gives. For $600 I can buy nearly 4 5870s!

[fpgaminer]

If you just want to fool around, start with this:
http://www.xilinx.com/products/boards-and-kits/AES-S6MB-LX9.htm

It's just $89.

Yeah you can get some cheap FPGA boards no problem. Terasic sells the DE0-Nano board for $79USD, or $59 if you're a student. That's for a 22K LUT device, which is pretty darn good for the price!

I'd respect you if you got 1 MH/s out of it.

You can get 1 MH/s out of it no problem.

[kokjo]

If you just want to fool around, start with this:
http://www.xilinx.com/products/boards-and-kits/AES-S6MB-LX9.htm

It's just $89.

Yeah you can get some cheap FPGA boards no problem. Terasic sells the DE0-Nano board for $79USD, or $59 if you're a student. That's for a 22K LUT device, which is pretty darn good for the price!

I'd respect you if you got 1 MH/s out of it.

You can get 1 MH/s out of it no problem.

can any FPGA work? how small can it be? will this be enogth: http://cgi.ebay.com/EP2C5T144-Altera-CycloneII-FPGA-mini-Development-Board-/270748203437?pt=BI_Electrical_Equipment_Tools&hash=item3f09d9b9ad#ht_1520wt_907
and how high can you clock the boards? without they are catching fire.

it would be nice to have a 1-2Mhash/s board, just for fun.

[Basiley]

This breaks my heart, open source FPGA mining code, yet the hardware is prohibitively expensive for the performance that it gives. For $600 I can buy nearly 4 5870s!

nowadays FPGA's had small density/per case[read small chips, less switches in package], made on obsolete waffer[0.13 in "best case" more usually 0.35], mean work on laughable frequencies and produced in small quatities primarily to IC designers, supers manufacturers and gov'ts.
when/if market grow up and/or someone invest in it[without which its unlikely happen], something may change.

[teknohog]

Compatible Board (and only purchase currently required):
Terasic DE2-115 Development Board

There must be cheaper and simpler boards with the same or equivalent FPGA. If you buy this board just for mining, all those nice I/O ports and displays are going to waste.

[allinvain]

This breaks my heart, open source FPGA mining code, yet the hardware is prohibitively expensive for the performance that it gives. For $600 I can buy nearly 4 5870s!

nowadays FPGA's had small density/per case[read small chips, less switches in package], made on obsolete waffer[0.13 in "best case" more usually 0.35], mean work on laughable frequencies and produced in small quatities primarily to IC designers, supers manufacturers and gov'ts.
when/if market grow up and/or someone invest in it[without which its unlikely happen], something may change.

I hope that happens, or it could also be the case that video cards will just get more and more powerful and they'll always dominate FPGAs. Where FPGA excell is at performance per watt, but not per $...if video cards double or triple their hashing performance but remain within the same thermal (PCIE 2.0 spec) envelope then GPU mining will still have a future.

Let's just hope that future ATI cards will remain excellent performers in doing straight integer work. But I have some doubs though as games are more floating point heavy and naturally graphics card manufacturers are going to optimize their architecture to perform better while gaming, not mining.

Imagine though if ATI caught wind of this whole mining phenomenon and they'd produce a Cayman core optimized for hashing!

[allinvain]

Compatible Board (and only purchase currently required):
Terasic DE2-115 Development Board

There must be cheaper and simpler boards with the same or equivalent FPGA. If you buy this board just for mining, all those nice I/O ports and displays are going to waste.

Hey let's form a mining group and approach the FPGA manufacturers and tell them if they can build for us a custom more stripped own version of their hardware we'd buy in bulk. I think aggregating purchasing power this way can cut costs down for everyone. Just a thought.