Minimalist Spartan6-LX150 board

[eldentyrell]

Updates: see this post and this one too.



Hi folks.  I've built a mining board for myself and I'm trying to
gauge interest to see if it's worth putting in a large PCB order and
reselling them (fully assembled).

Attached are photos of the version 7.0 board.  If all goes well I'll
be selling the version 8.0 boards, taking orders around 3-Oct and
shipping around 10-Oct.  The boards will sell for $250 in the basic
configuration, which is:

  - The board (85mm x 65mm)
  - Spartan6 LX150-2
  - Bypass capacitors
  - Xilinx JTAG connector
  - 10Mhz clock crystal
  - Mounting holes for a standard Northbridge heatsink (two different
    hole diameters; rotate the sink 90 degrees to change width).
  - High-current Molex connector (see below)

The black 20-pin right-angle connector shown in the photo will not be
on the final board, and some of the unsoldered pads will change or go
away.

The board is designed to take power from a backplane (see photo) using
a high-current Molex connector.  This connector is the same kind used
on ATX motherboards, but has a different pin count and keying so you
can't plug it in by accident (!).  The Molex connector carries 1.2V
(high current) and 3.3V power, ground, one I/O pin, and all the JTAG
wires.

The backplane (sold separately for $160) has slots for six boards and
includes a SATA power connector and 48W power supply.  The JTAG
chain enters via the backplane and snakes through all the boards, so
no extra cables are required.  Airflow is meant to come in through the
side.

Extra options (includes parts+assembly):

  - $15 for 12W power supply and SATA connector so you don't need the
    backplane.

  - $5 for a 20-pin connector carrying DONE, INIT_B, PROGRAM_B, ground,
    and 14 general-purpose I/O pins.  Or you can always solder this yourself later.
    This connector means that if you decide to stop mining, you
    can still use the board for plenty of other FPGA-related projects.

  - $20 to upgrade from a Spartan6 -2 to a -3, although I'm not convinced
    this is worth it.  Parts are often "graded" into the -2 bin for reasons that
    don't affect mining performance (I/Os and DSP slices).

Payment will be only via bitcoin; my margins on this are so slim that
it only takes one stolen-credit-card order to wipe out whatever
meager profit I'll make on two dozen boards.  Escrow is fine as long
as the buyer pays; I'm still trying to figure out which service(s) to
use.

Anyways, please let me know -- zero commitment -- how many boards you
would be likely to order at the pricing and availability dates above.
That will help me figure out if I should place a large PCB order or just get
enough for myself.

Thanks!

[1715582726]

1715582726

[1715582726]

1715582726

[1715582726]

1715582726

[1715582726]

1715582726

[thirdlight]

Interesting - what H/s are you getting?

[eldentyrell]

Interesting - what H/s are you getting?

The same as anybody else using a Spartan6-LX150.  There's nothing special about this board except for the backplane option and the fact that it costs a whole lot less than the boards sold by Digilent, etc.

I currently mine with my own HDL code -- which is not performing very well yet (!).  I have deliberately not looked at or downloaded fpgaminer's code in order to avoid being the target of accusations.

If somebody posts a bitstream for an LX150-2 with an 10Mhz clock input on pin B10 (IOSTANDARD=LVCMOS33), all other communication is over JTAG, I will be happy to try it out and post results here within a day or two.  Please include the timing report (twr) and map report (mrp).

In terms of testing the hardware itself, I have run a single "lane" of my own hashing code at 320mhz (which is the BRAM speed limit on a -3 and out-of-spec for a -2) and have run current draw tests (fill the chip with shift registers, crank up the clock speed) up to 10A using the backplane power supply.  Everything worked perfectly, no errors.

(FWIW, though, there's absolutely no hope of PAR finishing with the whole chip full of lanes at 320mhz, and I have to waste an ungodly quantity of registers on pipelining.  The point is that the power supply is stable and reliable at high clock speeds and high current draw, not that you'd actually want to do mining this way (you don't).)

[rph]

The ztex Verilog is GPL and can be easily modified to reach ~150MH/s in -3, with 2 cycles per SHA256 stage,
if you'd like to switch to that. See the discussion here https://bitcointalk.org/index.php?topic=40047.0

The through-hole bypass caps are just strange (WTF, don't those interfere with the heatsink?).
The price looks very competitive though.

-rph

[sadpandatech]

The ztex Verilog is GPL and can be easily modified to reach ~150MH/s in -3, with 2 cycles per SHA256 stage,
if you'd like to switch to that. See the discussion here https://bitcointalk.org/index.php?topic=40047.0

The through-hole bypass caps are just strange (WTF, don't those interfere with the heatsink?).
The price is very competitive though. Well done.

-rph

I was thinking the same thing about the heatsink on first sight. Only thing that makes sense is if the chip side of the heatsink has a smaller footprint than the outer edge we can see.

[eldentyrell]

The through-hole bypass caps are just strange

Keeps the cost down and yield up.  Everything except the BGA chip is assembled by hand, and hand-soldering SMT capacitors is notoriously unreliable/time-consuming.

(WTF, don't those interfere with the heatsink?)

The caps actually lay down on their sides (I took the first picture before doing this).  You also have to put a small 14mmx14mmx3mm "mini heatsink" (basically just a block of copper or aluminum) between the chip and the huge northbridge sink; I'll include this with any boards that are ordered since I'm sure I can get them for under a dollar.  Pressure from the springs on the northbridge keeps it in place, but you can always add a drop of tacky thermal compound if you're paranoid.

[Silverpike]

This is outstanding, I love it.  The power connector configuration is really great, I love the mini-backplane.

This is the nicest and most compact Spartan-6 board I have yet seen.  Great job!

[fivebells]

Will it be possible for the end-user to reprogram this thing?  Might be fun.

[Dexter770221]

48W from SATA connector thats over 4A of current from 12V rail. Specs are 1.5A from this rail. You may SMELL trouble if you populate all slots...

[Dargo]

I'm trying to get a ballpark estimate of the hash/s and wattage here. Each board consumes 12 watts (or less), and one might be able to get around 150 Mh/s out of it? Also, the advantage of the $160 back plane is entirely cosmetic (i.e. to nicely hold several boards in a compact way)?   

[Transisto]

48W from SATA connector thats over 4A of current from 12V rail. Specs are 1.5A from this rail. You may SMELL trouble if you populate all slots...

I sometime have to use four pci-e to molex, ~300w,  48w is nothing. What rail ? psu rails are 18A or more

[sadpandatech]

48W from SATA connector thats over 4A of current from 12V rail. Specs are 1.5A from this rail. You may SMELL trouble if you populate all slots...

I believe you are mixing up your rails. What you are likely refering to is this, 'Another ATX addition was the +5sb rail for providing a small amount of standby power at 1.5a, even when the computer was nominally "off".'  Even on the cheapest little power supply I could reach and grab real quick with a sata power connector on it, I find that it is on the same 12v rail as the other 4pin molex are. On this el cheapo 250w PSU the 12v rail is rated at 13a...

[bcforum]

I currently mine with my own HDL code -- which is not performing very well yet (!).  I have deliberately not looked at or downloaded fpgaminer's code in order to avoid being the target of accusations.

The legal definition of 'clean room software' is much more flexible than most people assume.

You can look at fpgaminer's (or any other available) design.
You can write detailed specifications based on the design while looking at it. You have to be careful not to include any copyrighted material in the specification (IE code snippets.)
The same person can use the specifications to develop a new design.

The only limitation is: you cannot reference the original code while developing the new code.

For everyone else, the LX150 isn't supported by the free version of the Xilinx tools. You'll have to buy the full version ($5K?) to develop for this board.

[Dexter770221]

What I had in my mind is that pins on connector can be damaged. Of course modern computer PSU can deliver much much more current but pins on connectors are too thin and can burn. What I've missed is the info that one pin can deliver 1.5A of current, there are 3 pins for +12V so this connector will be close to its limits.

[deslok]

What I had in my mind is that pins on connector can be damaged. Of course modern computer PSU can deliver much much more current but pins on connectors are too thin and can burn. What I've missed is the info that one pin can deliver 1.5A of current, there are 3 pins for +12V so this connector will be close to its limits.

a sata connector will be fine for that kind of ampload, i've got a few sat->6pin pcie connectors i've been using for months with no issue

[fivebells]

For everyone else, the LX150 isn't supported by the free version of the Xilinx tools. You'll have to buy the full version ($5K?) to develop for this board.

Thanks.

[norulezapply]

Definitely interested in this!

Any prices for shipping to the UK if it's available?

[eldentyrell]

Also, the advantage of the $160 back plane is entirely cosmetic (i.e. to nicely hold several boards in a compact way)?  

The backplane has the 48W DC-DC power supply on it.  You won't be able to do much without a power supply!  That's why there's the option to get a smaller 12W version soldered directly on to the spartan card -- you need that option if you don't get the backplane.

The backplane also daisy-chains the six boards' JTAG connections, so you only need ONE usb-to-JTAG adapter instead of six.

[eldentyrell]

I sometime have to use four pci-e to molex, ~300w,  48w is nothing. What rail ? psu rails are 18A or more

The 1.2V supply is stepped down off of the 12V rail; that's where 99% of the power is drawn.

A tiny amount of power is used at 3.3V; this comes from the 5V rail on the SATA connector.  It's less than 50mA; basically negligible.  Only used for JTAG, the clock input pad, and VCCAUX.

I don't use the SATA 3.3V pin because it isn't always connected (e.g. molex-to-SATA adapters).

[eldentyrell]

I currently mine with my own HDL code -- which is not performing very well yet (!).  I have deliberately not looked at or downloaded fpgaminer's code in order to avoid being the target of accusations.

The legal definition of 'clean room software' is much more flexible than most people assume.

It's not the legal definition I'm worried about -- it's reputation.

[eldentyrell]

Any prices for shipping to the UK if it's available?

I'm flexible on that; any insured service is fine.  By default I guess I'd just use the US Postal Service (which, I assume, does a handoff to the UK postal service somewhere before delivery).

However, buyer must pay any/all customs charges imposed by the receiving country.  You probably ought to fill out the customs form yourself and email it to me so I don't screw it up.  I'll set up a mechanism for doing this if there is enough interest to warrant a large run of PCB's.

[Dargo]

Also, the advantage of the $160 back plane is entirely cosmetic (i.e. to nicely hold several boards in a compact way)?  

The backplane has the 48W DC-DC power supply on it.  You won't be able to do much without a power supply!  That's why there's the option to get a smaller 12W version soldered directly on to the spartan card -- you need that option if you don't get the backplane.

I understand that, but $160/6 = $26.67 per board vs $15 for the 12W supply directly on the board. So I'm wondering if there is any advantage to the back plane to justify the higher cost, aside from the cosmetic point of having six boards neatly mounted on the back plane. Also, is 48W enough for six boards? That's only 8W per board.

Edit: Sorry, didn't see Big Chip's (added?) reply about JTAG connections - that answers my question.

[eldentyrell]

Also, is 48W enough for six boards? That's only 8W per board.

That's a good point.

Fortunately ArtForz pointed out a 72W (60A @ 1.2V) power supply that is schematic-compatible (though not pin-compatible) with the 48W one I've been using.  It's very likely that I'll switch.  This may increase the cost of the backplane by about $10.  Unfortunately they have different pinouts, so I can't offer a choice of one or the other.

One of the other reasons for choosing the backplane design is uncertainty about power requirements.  I have a hunch that FPGA mining will continue to increase the mhashes-per-lut, but from this point forward the increases will require increased power consumption (though still nothing close to GPUs).  So I'm trying to future-proof things a bit by separating the expensive chip from the power supply.

If some magical bitstream falls out of the sky that gives us twice the hashpower for twice the wattage, people can just leave half the slots on the backplane empty.  Sure those empty slots "cost" you $26 per board, but it's a lot better than having the power supply welded on to the board and being left out of the game completely!

[rph]

If some magical bitstream falls out of the sky that gives us twice the hashpower for twice the wattage

Heh, there's no way 6s150 will reach 300MH/s, even if you take several
months to hand-place and hand-route every LUT...

-rph

[Turbor]

Interested 

[eldentyrell]

If some magical bitstream falls out of the sky that gives us twice the hashpower for twice the wattage

Heh, there's no way 6s150 will reach 300MH/s, even if you take several
months to hand-place and hand-route every LUT...

-rph

Ah, you are underestimating the power of <blink>magical</blink> bitstreams!!!!

No seriously, the point is there will undoubtedly be ways to get (1+X)% the performance for (1+Y)% of the power, and I don't think anybody knows for sure right now what the X/Y values will be.

You're right that nobody's going to fit two whole copies of the current (6s150-sized) fpga miners on a 6s150 at any point in the future.  But there are less efficient mining cores that are much more flexibly routed and come at a much finer granularity.  This has been the course I've been pursuing lately with my own designs: find an inefficient-yet-better-than-nothing use for the idle LUTs left behind after PARing the ordinary design.

[ultraimports]

I am veeerrry interested, pausing my own FPGA purchase to see where this goes....keep me updated (sub'd).

[nelisky]

How much power would the 6 boards + backplane take? Expecting 900MHs out of it, and if you sell these plug & mine, I'll almost certainly be in for on full backplane for testing!

[322i0n]

How much power would the 6 boards + backplane take? Expecting 900MHs out of it, and if you sell these plug & mine, I'll almost certainly be in for on full backplane for testing!

Me too.

[Philj]

I'm interested.

What would be required to get one device connected to a standard PC? The base device($250), plus a single 12w power suply ($15)? Are you including any connector/cable to get I/O to/from the PC, or would an additional cable (jtag to usb???) be required?

Just trying to get an idea of just how much would be needed to attach this to a basic PC. I don't want to budget for $265 per card then find out that each card also needs a $50 cable or something like that to make it actually work.

[dovewing2000]

notified - though i'm still wondering what the speed will be like...

[eldentyrell]

How much power would the 6 boards + backplane take?

That depends on what bitstream you use.

Most of the posts I've seen on this site quote under 7W per board, but like I've said before I believe that will go up in future designs.  Anyways, power consumption is determined by the chip and the bitstream (unless the board designer did something really stupid) so you can assume this board's power consumption will be pretty much the same as any other.

[eldentyrell]

What would be required to get one device connected to a standard PC? The base device($250), plus a single 12w power suply ($15)? Are you including any connector/cable to get I/O to/from the PC, or would an additional cable (jtag to usb???) be required?

Just trying to get an idea of just how much would be needed to attach this to a basic PC. I don't want to budget for $265 per card then find out that each card also needs a $50 cable or something like that to make it actually work.

You need a USB-to-JTAG cable.  One of the important features of the backplane is that it lets you use ONE cable for six boards, so you get to split the cost of the cable six ways.

Some bitstreams require that you use Xilinx's lame IMPACT software and an IMPACT-compatible JTAG cable (they tend to be more expensive).

I personally use urjtag in my own (unreleased) designs, and there are $20 cables that work with it.  But I don't think the "fpgaminer" and "ztex" designs work with them.

[teek]

What would be required to get one device connected to a standard PC? The base device($250), plus a single 12w power suply ($15)? Are you including any connector/cable to get I/O to/from the PC, or would an additional cable (jtag to usb???) be required?

Just trying to get an idea of just how much would be needed to attach this to a basic PC. I don't want to budget for $265 per card then find out that each card also needs a $50 cable or something like that to make it actually work.

You need a USB-to-JTAG cable.  One of the important features of the backplane is that it lets you use ONE cable for six boards, so you get to split the cost of the cable six ways.

Some bitstreams require that you use Xilinx's lame IMPACT software and an IMPACT-compatible JTAG cable (they tend to be more expensive).

I personally use urjtag in my own (unreleased) designs, and there are $20 cables that work with it.  But I don't think the "fpgaminer" and "ztex" designs work with them.

Are you selling the cable with the units? 

[eldentyrell]

Are you selling the cable with the units? 

No.

[eldentyrell]

Well, seems like quite a lot of people are interested in this thread, but so far only one posting from somebody who actually mentions being interested in buying a board.

Looks like I will only be ordering enough PCBs for my own personal use.

This isn't a big deal for me because I designed the boards mainly for myself, figuring the ability to sell them would be neat if it worked out -- but not depending on it.

However, for anybody out there thinking of designing/building/marketing a Spartan6 board for mining purposes, it looks like the market isn't interested unless the fully-assembled cost is below $250 per chip.  Probably an important data point to have before you make any plans for large-scale production!

[teek]

Are you selling the cable with the units? 

No.

Could you provide a link to the cable?  Might be nice if you offered the cable you use for an additional fee..

[teek]

Well, seems like quite a lot of people are interested in this thread, but so far only one posting from somebody who actually mentions being interested in buying a board.

Looks like I will only be ordering enough PCBs for my own personal use.

This isn't a big deal for me because I designed the boards mainly for myself, figuring the ability to sell them would be neat if it worked out -- but not depending on it.

However, for anybody out there thinking of designing/building/marketing a Spartan6 board for mining purposes, it looks like the market isn't interested unless the fully-assembled cost is below $250 per chip.  Probably an important data point to have before you make any plans for large-scale production!

i'm quite interested in buying some boards..

[fizzisist]

Well, seems like quite a lot of people are interested in this thread, but so far only one posting from somebody who actually mentions being interested in buying a board.

Looks like I will only be ordering enough PCBs for my own personal use.

This isn't a big deal for me because I designed the boards mainly for myself, figuring the ability to sell them would be neat if it worked out -- but not depending on it.

However, for anybody out there thinking of designing/building/marketing a Spartan6 board for mining purposes, it looks like the market isn't interested unless the fully-assembled cost is below $250 per chip.  Probably an important data point to have before you make any plans for large-scale production!

It makes me happy to see more designs out there. Competition is a good thing! Especially, if others are developing different designs that will suit different needs. This backplane concept looks very nice.

That said, I don't think the cost advantage of your board over, say, ours is as large as it might appear. Our boards have the -3 speed grade by default, so we should compare the $270 price. Fully assembled with 6 boards, your system costs 6*270+160+20 = $1800. Our boards cost $610, so an equivalent system (3*X6500) would be $1830 (plus three USB cables). Also, do your boards include heatsinks?

As for the design, those through-hole capacitors are troubling. Take a look at UG393. They spend pages stressing the importance of capacitor package size and placement, even taking the time to explain a lot of fundamental PCB design concepts. Definitely worth a read.

[eldentyrell]

However, for anybody out there thinking of designing/building/marketing a Spartan6 board for mining purposes, it looks like the market isn't interested unless the fully-assembled cost is below $250 per chip.  Probably an important data point to have before you make any plans for large-scale production!

That said, I don't think the cost advantage of your board over, say, ours is as large as it might appear.

Perhaps.  My point is that even this price point is not low enough, so you guys might want to hold off on high-volume production until you can go even lower.  Apparently neither of us is at a low enough price point.

As for the design, those through-hole capacitors are troubling. Take a look at UG393. They spend pages stressing the importance of capacitor package size and placement, even taking the time to explain a lot of fundamental PCB design concepts. Definitely worth a read.

Yes, I read it many times... did you?  There's no prohibition against through-hole caps and they explicitly say that it's okay to have the 0.47uF's outside the footprint as long as they're within half an inch of the device outline.

[Turbor]

Well for people like me (carpenter) all this technical things are a bit hard to understand. The prize is one thing but for me it's more important to understand first how it works. I need a kit i can plug in and mine. The total package. When you can offer that (with enough hash rate), i'm sure that people will buy.

[322i0n]

I need a kit i can plug in and mine. The total package. When you can offer that (with enough hash rate), i'm sure that people will buy.

agree this

[newMeat1]

This is phenomenal that you could get the cost down so low. My only complaint was going to be the JTAG interface. But, since you've accounted for daisy-chaining 6 of them... that complaint is gone. Hopefully your JTAG signal integrity holds up from daisy-chaining 6 boards. It's touchy.

Give the bitcoin community some time to warm up to your idea. There've been enough scams going around, so it's no surprise they don't jump on the bandwagon right away.

About fpgaminer's open-source code and ztex's "borrowing" of it-- the issue wasn't that ztex borrowed a few bits and pieces. It is open-source after all. I think the issue was lack of acknowledgement.

[fizzisist]

Yes, I read it many times... did you?  There's no prohibition against through-hole caps and they explicitly say that it's okay to have the 0.47uF's outside the footprint as long as they're within half an inch of the device outline.

Hmm, I disagree... Anyway, your design works, so that's good! It just might not be as immune to noise as it would be with 0402s. It's not a mortal sin to disobey the datasheet, just risky.

[jjiimm_64]

I need a kit i can plug in and mine. The total package. When you can offer that (with enough hash rate), i'm sure that people will buy.

Indeed.

I am with these guys..  I am no electrical engineer..  but I can plug in a psu

[rph]

Once miners start posting success stories and pics of stable, large, $1-2/MH FPGA rigs,
the demand will follow. Anybody buying today would be a super-early-adopter.
(Which I guess makes ArtForz a super-super-super-early adopter...)

It's a huge PITA to build and cool a 100 GPU rig, but a 200-300+ FPGA rig is no problem in a tiny apartment.
Without dummy monitor plugs, $500 of molex<->PCI-e cables, 50 ATX power supplies, and all the other non-wife-approved stuff.

Once there are enough FPGAs on the network, difficulty will increase and GPUs will become unprofitable or barely
profitable for anyone paying for cooling + electricity [probably most people with more than 4-5 GPUs]. It's a self-fulfilling prophesy.
Live or die, make your choice 

-rph

[ngzhang]

Once more miners start posting success stories and pictures of stable, large, $1-2/MH FPGA rigs,
the demand will follow. Anybody buying today would be a super-early-adopter.
(Which I guess makes ArtForz a super-super-super-early adopter...)

It's a huge PITA to build and cool a 100 GPU rig, but a 200-300+ FPGA rig is no problem in a tiny apartment.
Without dummy monitor plugs, $500 of molex<->PCI-e cables, 50 ATX power supplies, and all the other non-wife-approved stuff.

Once there are enough FPGAs on the network, difficulty will increase and GPUs will become unprofitable or barely
profitable for anyone paying for cooling + electricity [probably most people with more than 4-5 GPUs]. It's a self-fulfilling prophesy.
Live or die, make your choice  

-rph

These situation will or not come to truth depends on YOU HDL programmers and custom FPGA board venders.
My DUAL spartan6-SLX150 board for development will small-lot on offer in 1week. I think a 400MH/s perboard(200 per FPGA, and certainly MUST with yours MAGIC bitstream) hashing speed with a price less than 600$ will attract more people turn to FPGA mining.

[rph]

Well, with the price drop, the payoff for GPUs is bleak already.
A completely DIY 6s150 should pay for itself faster than the "best" ATI option,
if you agree with the Bitminer math.

The exact #s are endlessly debatable of course (and not everybody will DIY).
[And some people live directly next to a hydro dam, or get free power through
various legal and illegal ways..]

Still I think FPGAs have already overtaken GPUs for the guys with real $$$$
to invest in mining. And, if not, at least they will very soon.

-rph

[Desolator]

I think there would be more customers if you translated some of this info from electronics expert into computer expert I have about enough electronics experience to know about electricity and ummm...well, that looks an awful lot like 3 liquid looking caps in your pic which are going to explode over the not-so-long long term if they're exposed to heat constantly You know, like cheap liquid caps that some uncaring manufacturer like HP mounts a quarter inch from an old Pentium 4 that leak after 1 years of significant enough dust build up as opposed to modern solid capacitors on ASUS boards that you could basically run in an oven for 10 years.  Get some of those 50,000 MTBF Japanese ones

That concern aside, everyone keeps making models and potential products and asking for preorders without ever explaining to us computer people the process outside the board that we're familiar with.  In other words, what's the hookup style (USB?), who wrote a driver if anyone, what versions of windows would it operate correctly on, and what mining software recognizes it as a processing device and can send and receive data for mining?  I bet you'd get 10x the interest if you made a quick post outlining all that.

[eldentyrell]

This is phenomenal that you could get the cost down so low. My only complaint was going to be the JTAG interface. But, since you've accounted for daisy-chaining 6 of them... that complaint is gone. Hopefully your JTAG signal integrity holds up from daisy-chaining 6 boards. It's touchy.

You raise an important point, although you can always crank the JTAG clock rate down to something ridiculous like 0.25Mhz.  Sure, it will take a full minute or two to load the bitstream, but once that's done you've still got an order of magnitude more bandwidth than you need for loading getwork's and reading back nonces.

So, yes, signal integrity is an issue here and I fully expect that the JTAG clock rate will need to be reduced when all six slots are full.  Once I have another four boards built I will know this for sure.  But the reduced clock rate will not affect mining throughput at all -- just "boot up" time.

[eldentyrell]

well, that looks an awful lot like 3 liquid looking caps in your pic which are going to explode over the not-so-long long term if they're exposed to heat constantly You know, like cheap liquid caps that some uncaring manufacturer like HP mounts a quarter inch from an old Pentium 4 that leak after 1 years of significant enough dust build up as opposed to modern solid capacitors on ASUS boards that you could basically run in an oven for 10 years.  Get some of those 50,000 MTBF Japanese ones

That's actually a pretty good idea and won't even require a PCB change.  Spending an extra buck or two on something non-electrolytic is probably worthwhile insurance against capacitor plague.  Thanks for the suggestion!

[Desolator]

well, that looks an awful lot like 3 liquid looking caps in your pic which are going to explode over the not-so-long long term if they're exposed to heat constantly You know, like cheap liquid caps that some uncaring manufacturer like HP mounts a quarter inch from an old Pentium 4 that leak after 1 years of significant enough dust build up as opposed to modern solid capacitors on ASUS boards that you could basically run in an oven for 10 years.  Get some of those 50,000 MTBF Japanese ones

That's actually a pretty good idea and won't even require a PCB change.  Spending an extra buck or two on something non-electrolytic is probably worthwhile insurance against capacitor plague.  Thanks for the suggestion!

hehehehehehe I soooo want to add a "see also: emachines.com" or maybe foxconn to the end of that article for no reason and see if anyone notices   Didn't know they named it btw.  I've seen at least a dozen myself with my small PC repair business.  One time I even saw one out in the middle of nowhere on the board bulge and fail causing the onboard graphics to distort and fail to initialize almost 100% of the time and it wasn't exposed to much heat.  So don't get cheap caps either apparently I'm sure there are plenty of high failure rate solid capacitor knock offs out there given their current reputation as being awesome.

[rph]

Please consider SMD ceramics..

-rph

[eldentyrell]

I need a kit i can plug in and mine. The total package. When you can offer that (with enough hash rate), i'm sure that people will buy.

Indeed.

I am with these guys..  I am no electrical engineer..  but I can plug in a psu

You have a good point.  But to flip that around, on the GPU side it's taken as a given that you're going to buy your hardware from a company (ATI) that does not provide the mining software (or even admit it knows what bitcoin is).  But I understand that while gamers have seen GPUs before, most bitcoiners are encountering FPGAs for the first time.  They aren't scary; they're just obscenely flexible... "enough rope to hang yourself with."

I could, perhaps, put together a turn-key solution, although it would involve a lot of effort.  My two major concerns are:

1. HDL developer guilt.  It makes me slightly ill to see posts like ngzhang's "hey you lazy-ass HDL developers make your code faster so I can make MOAR PROFITZ!!!".  I'd feel queasy about selling a "solution" that bundled in somebody else's hard work.  I don't know the exact details of the fpgaminer/ztex dispute, but I can certainly empathize with the initial reaction from fpgaminer.  It would make me really happy to be providing low-cost boards to people who are interested in tweaking/tuning/improving the HDL code, but I think I've figured out now that there aren't as many of those people as I'd thought.

2. Support.  I'm happy to help out here in a casual message-board-member way.  But I'm kinda worried about lazy users buying a "turn-key" solution from me and then demanding that I hand-hold them through the whole process of configuring Xilinx's crapware drivers on their Windows host box (I haven't used Windows in almost a decade) under threat of posting negative reviews of my product ("did not work for me").  I definitely can't sell the boards for $250 if I have to budget in my own time spent on extensive tech support work.

Anyways.  Looks like the first run will be small personal-use-only, but there may be another batch of boards in November after I've figured out if it's worth taking this to the next level.

[eldentyrell]

Please consider SMD ceramics..
-rph

I would if I owned a (tens-of-thousands-of-dollars) pick-and-place machine.  Unfortunately trying to do those by hand is painful and unreliable.  I just can't justify it on an eyestrain/benefit basis.

[ngzhang]

I need a kit i can plug in and mine. The total package. When you can offer that (with enough hash rate), i'm sure that people will buy.

Indeed.

I am with these guys..  I am no electrical engineer..  but I can plug in a psu

You have a good point.  But to flip that around, on the GPU side it's taken as a given that you're going to buy your hardware from a company (ATI) that does not provide the mining software (or even admit it knows what bitcoin is).  But I understand that while gamers have seen GPUs before, most bitcoiners are encountering FPGAs for the first time.  They aren't scary; they're just obscenely flexible... "enough rope to hang yourself with."

I could, perhaps, put together a turn-key solution, although it would involve a lot of effort.  My two major concerns are:

1. HDL developer guilt.  It makes me slightly ill to see posts like ngzhang's "hey you lazy-ass HDL developers make your code faster so I can make MOAR PROFITZ!!!".  I'd feel queasy about selling a "solution" that bundled in somebody else's hard work.  I don't know the exact details of the fpgaminer/ztex dispute, but I can certainly empathize with the initial reaction from fpgaminer.  It would make me really happy to be providing low-cost boards to people who are interested in tweaking/tuning/improving the HDL code, but I think I've figured out now that there aren't as many of those people as I'd thought.

2. Support.  I'm happy to help out here in a casual message-board-member way.  But I'm kinda worried about lazy users buying a "turn-key" solution from me and then demanding that I hand-hold them through the whole process of configuring Xilinx's crapware drivers on their Windows host box (I haven't used Windows in almost a decade) under threat of posting negative reviews of my product ("did not work for me").  I definitely can't sell the boards for $250 if I have to budget in my own time spent on extensive tech support work.

Anyways.  Looks like the first run will be small personal-use-only, but there may be another batch of boards in November after I've figured out if it's worth taking this to the next level.

Here we have a team working hard on the HDLs, but unfortunately, it's really a extreme difficult work.

[P4man]

I'd feel queasy about selling a "solution" that bundled in somebody else's hard work. 

Then make a deal with the someone else's and give them a share of the revenue (possibly in return for them providing support on the software)

2. Support.  I'm happy to help out here in a casual message-board-member way.  But I'm kinda worried about lazy users buying a "turn-key" solution from me and then demanding that I hand-hold them through the whole process of configuring Xilinx's crapware drivers on their Windows host box

No harm in stating you'd only support Linux. I think most serious miners use linux anyway. At least the market for "linux able" miners is infinitely bigger than the market for people who are familiar with FPGAs.

Just my 2 cents.

[sirky]

Following.

[kripz]

Interested

[DeathAndTaxes]

Once there are enough FPGAs on the network, difficulty will increase and GPUs will become unprofitable or barely
profitable for anyone paying for cooling + electricity [probably most people with more than 4-5 GPUs]. It's a self-fulfilling prophesy.

I often see this quoted but it is nonsense.  Higher difficulty will make FPGA even @ $2 per MH even MORE prohibitively expensive.  Higher difficulty benefits those w/ efficient GPU (like 5970 & 7xxx series) and moderate to low cost electricity the most.

I think you will see a difficulty spike will kill demand for new FPGA not drive it.

Take hypothetical FPGA miner $2 per MH.  150MH = $300 in cost.  Running 24/7/365 @ 15W.
Break even @ current difficulty is 25 months.
Break even @ 30% difficulty increase is 33 months.
Break even @ 50% difficulty increase is 40 months.

Currently today one could buy 5970 for <$500.  Say 3x5970 + powersupply + other components = 2.8GH for $2800.  Running 24/7/365 @ 1000W.

Break even @ current difficulty is 17 months.
Break even @ 30% difficulty increase is 25 months.
Break even @ 50% difficulty increase is 32 months.

Difficulty increases close the gap but even $2 per MH (an impressive improvement) is still undercut by anyone w/ $0.10 electrical costs (or less). I am interested in FPGA but these dire predictions of them killing GPU are simply unwarranted unless cost is closer to $1 per MH installed.

Remember GPU performance per watt won't be static.  The 7xxx series looks to almost double performance per watt (cutting electrical cost in half for GPU miners).  A break even of 40+ months, is a considerable risk as 4 years is long enough for 2 product cycles in GPU world.  The product after the 7xxx series likely won't improve performance per watt (think a repeat of 5xxx vs 6xxx) but the generation after that (lets call it 9xxx series) will be a move to 20nm and bring all the power reduction and performance boosts that a die shrink does.

4 years is a long time.  My comparison above is based on the 5970s.  Soon FPGA will compete against 7xxx series (nearly double the performance per watt) and within 4 years against the 9xxx series (4x the performance per watt).

[ultraimports]

Once there are enough FPGAs on the network, difficulty will increase and GPUs will become unprofitable or barely
profitable for anyone paying for cooling + electricity [probably most people with more than 4-5 GPUs]. It's a self-fulfilling prophesy.

I often see this quoted but it is nonsense.  Higher difficulty will make FPGA even @ $2 per MH even MORE prohibitively expensive.  Higher difficulty benefits those w/ efficient GPU (like 5970 & 7xxx series) and moderate to low cost electricity the most.

I think you will see a difficulty spike will kill demand for new FPGA not drive it.

Take hypothetical FPGA miner $2 per MH.  150MH = $300 in cost.  Running 24/7/365 @ 15W.
Break even @ current difficulty is 25 months.
Break even @ 30% difficulty increase is 33 months.
Break even @ 50% difficulty increase is 40 months.

Currently today one could buy 5970 for <$500.  Say 3x5970 + powersupply + other components = 2.8GH for $2800.  Running 24/7/365 @ 1000W.

Break even @ current difficulty is 17 months.
Break even @ 30% difficulty increase is 25 months.
Break even @ 50% difficulty increase is 32 months.

Difficulty increases close the gap but $2 per MH is still beat by anyone w/ $0.10 electrical costs (or less).  I am interested in FPGA but these dire predictions of them killing GPU are simply unwarranted unless cost is closer to $1 per MH installed.

Remember GPU performance per watt won't be static.  The 7xxx series looks to almost double performance per watt (cutting electrical cost in half for GPU miners).  A break even of 40+ months is highly dangerous one risks being undercut by the next next gen video cards.  4 years is long enough for 2 product cycles and we will be looking @ 20nm chips (and other doubling of performance per watt).

Very good info. I guess I would ask the FPGA experts: how often do the FPGA chips increase in performance as well? Do they move as fast as GPUs? Do they follow Moore's law, essentially?

[P4man]

Im not an FPGA  expert by any stretch but they definitely follow Moore's law, in fact arguably more easily than CPU's, as they are much simpler and you can just up the # units as your process gets smaller.  Similar to how, because CPU designs hit an IPC  and clock scaling brick wall, most of the extra transistor budget is simply spent on going from single to dual, quad and octal cores.

That said, Im not sure I agree with the above math; the market dynamic of mining leads to difficulty gravitating  towards break even point for the average miner. I suspect most miners see their hardware investment as sunk cost, leaving the electricity bill. FPGA's already have better MH/Watt and I suspect that gap will grow as the software matures. Its true its a risky investment, certainly at this point, but once this starts generating sufficient volume I can see prices tumbling. after all, an FPGA is likely cheaper to produce than our highend gaming GPUs.

[DeathAndTaxes]

Its true its a risky investment, certainly at this point, but once this starts generating sufficient volume I can see prices tumbling. after all, an FPGA is likely cheaper to produce than our highend gaming GPUs.

Most of the cost (60%+) comes from the actual FPGA.  It is unlikely prices will tumble.  FPGA already have economies of scale.  20 million are sold each year.  Another 1K (or even 10K) miners using FPGA isn't going to cause a massive drop in price. Maybe if one of the FPGA developers gets a massive buy order they could cut FPGA & assembly costs by 30%.  Software improvements might squeeze out another 10%-20% out of current gen FPGA but that still only gets us to ~$2/MH installed.

Yes FPGA benefit from Moore's law but so will GPU.  GPU are almost perfectly scalable.  When the process size gets cut in half then you simply double the numbers of shaders, get roughly 2x the performance and the die size (and thus cost & power) remains the same.

I have derailed this thread enough as it is.  To the OP very good work looks promising I just find the false hope of people (not you) pretending away the economic issues of FPGA (and claims of the death of GPU mining) to be naive & frustrating.

[vx609e]

Following

[rph]

I just find .. claims of the death of GPU mining to be naive & frustrating.

I've invested a lot of time into FPGA mining. Here is my thinking:

If GPUs remain the dominant technology, difficulty will adjust to make them barely profitable in average-electricity-cost areas.
I don't think anyone really disagrees with that, it's an intentional design decision in Bitcoin.

Once that happens: GPUs in high-elec-cost areas (like me) will be unprofitable. FPGAs will be profitable everywhere operationally
[in terms of BTC produced minus electricity/cooling/maintenance costs]. So they will eventually pay for themselves,
unless Bitcoin collapses entirely first, screwing over all miners. The payoff might take 2 years, but that is still a pretty decent ROI compared
to FDIC savings accounts, the stock market, treasuries, etc. My FPGAs won't lose 30% overnight due to some Goldman Sachs bullshit.

If/when 28nm or 20nm or 16nm GPUs are driving the difficulty, my 45nm FPGAs will still have better MH per watt, so they will
still be profitable operationally. And anyway I will then be adding 28nm or 20nm or 16nm FPGAs.

If FPGAs become the dominant technology, difficulty will adjust to make them barely profitable in average-power-cost areas of the world.
GPUs will then be wildly unprofitable everywhere, except for people that somehow have free electricity
[which I think is a tiny fraction of the network]. Then we'll see $50 5830s on eBay as lots of people rush to the exits.

I actually hope that GPUs remain the dominant technology, while I mine on FPGAs, with a nice, high profit margin.

If a very high-end ASIC becomes the dominant technology then both GPUs + FPGAs will be unprofitable operationally.
I seriously doubt this will happen. The people with the skills and capital to make it happen could make a lot more money
with less risk building something else. [I'm not talking about a Mosis 250nm ASIC; I'm talking 90nm or better]

that still only gets FPGAs to ~$2/MH installed.

$1/MH is possible today if you build the boards yourself, or professionally in qty 100+

I suspect most miners see their hardware investment as sunk cost, leaving the electricity bill. FPGA's already have better MH/Watt and I suspect that gap will grow as the software matures.

Exactly. The decision to keep a GPU running, or shut it off, is not based on some breakeven calculation you did when you bought it.
It's based on whether it's making or losing money, today, based on current difficulty + elec/cooling costs.
I stand by my statement that if FPGAs take off, they will certainly put most GPU miners out of business,
and capture a large percentage of the coins left to be mined.

-rph

[ngzhang]

I just find .. claims of the death of GPU mining to be naive & frustrating.

I've invested a lot of time into FPGA mining. Here is my thinking:

If GPUs remain the dominant technology, difficulty will adjust to make them barely profitable in average-electricity-cost areas of the world.
I don't think anyone really disagrees with that, it's an intentional design decision in Bitcoin.

Once that happens: GPUs in high-elec-cost areas (like me) will be unprofitable. FPGAs will be profitable everywhere operationally
[in terms of BTC produced minus electricity/cooling/maintenance costs]. So they will eventually pay for themselves,
unless Bitcoin collapses entirely before then, screwing over all miners. It might take 2 years, but that is still a pretty decent ROI compared
to FDIC savings accounts, the stock market, treasuries, etc.

This is true even if, say, 28nm or 20nm GPUs are driving the difficulty. My 45nm FPGAs will still have better MH per watt, so they will
still be profitable operationally.

If FPGAs become the dominant technology, difficulty will adjust to make them barely profitable in average-power-cost areas of the world.
GPUs will then be wildly unprofitable everywhere, except for people that somehow have free electricity
[which I think is a tiny fraction of the network].

I actually hope that GPUs remain the dominant technology, while I mine on FPGAs, with a nice, high profit margin.

that still only gets FPGAs to ~$2/MH installed.

$1/MH is possible today if you build the boards yourself, or professionally in qty 100+

I suspect most miners see their hardware investment as sunk cost, leaving the electricity bill. FPGA's already have better MH/Watt and I suspect that gap will grow as the software matures.

Exactly. The decision to keep a GPU running, or shut it off, is not based on some breakeven calculation you did when you bought it.
It's based on whether it's making or losing money, today, based on current difficulty + elec/cooling costs.
I stand by my statement that if FPGAs take off, they will certainly put most GPU miners out of business,
and capture a large percentage of the coins left to be mined.

-rph

In a short time, 1.5$/MH will come true by some "low-manufacture-cost-area" of the world. by my troth.
Because the FPGA mining system is really not came to a big business, so I think profession groups still not join this game. In fact, miners are still a very small group. At this time, most of people making hard work on FPGA mining system maybe really for their "love". Their technique and effort surely could make more money in other filed.

[fivebells]

ngzhang, I am pretty much clueless about hardware, so I am interested in your views.  Don't you think that if a large-scale enterprise were to get into this, they would be more interested in making a custom ASIC than an FPGA?  How substantial do you imagine the power/speed gains could be for an ASIC over a GPU?

[ngzhang]

ngzhang, I am pretty much clueless about hardware, so I am interested in your views.  Don't you think that if a large-scale enterprise were to get into this, they would be more interested in making a custom ASIC than an FPGA?  How substantial do you imagine the power/speed gains could be for an ASIC over a GPU?

In my opinion, none of company will take part in any mining ASICs. If they have enough resource to tape out an ASIC, I'm sure they will design a another project, not for mining. As a single person, we can behaviour by interests, but a real company can't.

And answer your question, 1/10 cost, 10X performance, 1/10 energy consumption, on a single ASIC. AT LEAST.

[P4man]

In my opinion, none of company will take part in any mining ASICs.

Not specific for bitcoin, but SHA256 has other uses. VIA has CPU's with hardware accelerated encryption functions, and I thought recent (or upcoming?) Intel chips did do. They are no match for GPUs but it shows its already done. Also when I google "SHA256 chip" you find among  others, this:
http://www.s2cinc.com/product/pd.asp?id=278

I have no clue how that performs compared to our GPUs, or even if its usable for bitcoin mining, but I would be surprised if there werent chips out there or coming that could be used for bitcoin, even if they are not designed for bitcoin.

[ngzhang]

In my opinion, none of company will take part in any mining ASICs.

Not specific for bitcoin, but SHA256 has other uses. VIA has CPU's with hardware accelerated encryption functions, and I thought recent (or upcoming?) Intel chips did do. They are no match for GPUs but it shows its already done. Also when I google "SHA256 chip" you find among  others, this:
http://www.s2cinc.com/product/pd.asp?id=278

I have no clue how that performs compared to our GPUs, or even if its usable for bitcoin mining, but I would be surprised if there werent chips out there or coming that could be used for bitcoin, even if they are not designed for bitcoin.

there are some tiny difference between standard SHA256 hashing and bitcoin hashing. So ...

[eldentyrell]

"My FPGAs won't lose 30% overnight due to some Goldman Sachs bullshit."

I am tempted to make this my new .signature

[gopher]

I need a kit i can plug in and mine. The total package. When you can offer that (with enough hash rate), i'm sure that people will buy.

Indeed.

I am with these guys..  I am no electrical engineer..  but I can plug in a psu

You have a good point.  But to flip that around, on the GPU side it's taken as a given that you're going to buy your hardware from a company (ATI) that does not provide the mining software (or even admit it knows what bitcoin is).  But I understand that while gamers have seen GPUs before, most bitcoiners are encountering FPGAs for the first time.  They aren't scary; they're just obscenely flexible... "enough rope to hang yourself with."

I could, perhaps, put together a turn-key solution, although it would involve a lot of effort.  My two major concerns are:

1. HDL developer guilt.  It makes me slightly ill to see posts like ngzhang's "hey you lazy-ass HDL developers make your code faster so I can make MOAR PROFITZ!!!".  I'd feel queasy about selling a "solution" that bundled in somebody else's hard work.  I don't know the exact details of the fpgaminer/ztex dispute, but I can certainly empathize with the initial reaction from fpgaminer.  It would make me really happy to be providing low-cost boards to people who are interested in tweaking/tuning/improving the HDL code, but I think I've figured out now that there aren't as many of those people as I'd thought.

2. Support.  I'm happy to help out here in a casual message-board-member way.  But I'm kinda worried about lazy users buying a "turn-key" solution from me and then demanding that I hand-hold them through the whole process of configuring Xilinx's crapware drivers on their Windows host box (I haven't used Windows in almost a decade) under threat of posting negative reviews of my product ("did not work for me").  I definitely can't sell the boards for $250 if I have to budget in my own time spent on extensive tech support work.

Anyways.  Looks like the first run will be small personal-use-only, but there may be another batch of boards in November after I've figured out if it's worth taking this to the next level.

Hi big-chip-small-board,

I have been a lurker here for some time, more importantly a huge fan of your work.

May I offer you my opinion on this tricky matter.

I think that you should explore the advantage of your HDL developer quitting and streamline/optimise your work, in other words re-position it elegantly so it gets better accepted among the broader bitcoin community base.

The problems I see arise from the fact that FPGA developers try to tackle both parts of this project - design the hardware as well as develop the core functionality.

You know that this particular project requires two fundamentally different resources, one that is averagely good in both both areas will not do.

Hence, what about re-looking your marvellous hardware concept, perhaps give it a final touch and ensure that it is 100% compatible with the open source core functionality or any other firmware there is, so the people who buy your hardware can decide themselves what to run on it.

Instead of trying to excel in two areas, you then need to excel in one - design and produce great hardware - and let others develop the software and add value as they can.

This approach is very similar to the time when the GPU mining started - everyone would agree that selection of hardware and software are interlinked, but they got it from different sources.

Like a lot of people will say NVIDIA is capable of GPU mining, but there are not that many CUDA code written - because the hardware is not that good, or who know, because the god software developers only could afford to have ATI hardware to focus their development on.

But the result is clear - at the end of the day, ATI does not write hashing code and the bitcoin script writers do not develop highly-integrated hardware - and irrespective of this division of functionalities, the bitcoin community does not have problem locating the required bits and pieces and building their rigs.

My 2 cents.

[DeathAndTaxes]

Like a lot of people will say NVIDIA is capable of GPU mining, but there are not that many CUDA code written - because the hardware is not that good, or who know, because the god software developers only could afford to have ATI hardware to focus their development on.

Please tell me you are kidding.
1) You are aware many of the software developers do this full time as their day job.  I am sure someone with $50K to $120K salary can afford an NVidia card.
2) That has absolutely nothing to do with why Nvidia performance is so poor.

[gopher]

Like a lot of people will say NVIDIA is capable of GPU mining, but there are not that many CUDA code written - because the hardware is not that good, or who know, because the god software developers only could afford to have ATI hardware to focus their development on.

Please tell me you are kidding.
1) You are aware many of the software developers do this full time as their day job.  I am sure someone with $50K to $120K salary can afford an NVidia card.
2) That has absolutely nothing to do with why Nvidia performance is so poor.

You are missing my point - I have no idea why the developers have not develop good CUDA code - I only speculated to one of the possible reasons.

[P4man]

You are missing my point - I have no idea why the developers have not develop good CUDA code - I only speculated to one of the possible reasons.

I have no reason to believe the cuda miners arent any good. Its the nvidia cards that arent as suited to bitcoin mining as amd cards, due to fundamentally different architectures. Since that makes the cards uncompetitive, it stands to reason few people will invest heavily in cuda apps that can only work on this (for bitcoin) very uncompetitive hardware. No amount of software optimization is going to turn a 140 MH/s nVidia card in to a 400 one. There is probably less than 10% untapped potential.

[sirky]

This is all true. Just look at the shader counts between NVidia and AMD cards and you have your answer. The processors (shaders) have to do the work, and NVidia cards don't have as many.

[DeathAndTaxes]

You are missing my point - I have no idea why the developers have not develop good CUDA code - I only speculated to one of the possible reasons.

You are missing the point.  There are VERY GOOD CUDA miners.  It is unlikely any future CUDA miner would get more than 10% more performance out of existing cards.

Nvidia hardware just happens to be ill-suited for integer math (the math used in hashing).

[DeathAndTaxes]

Not speaking on the architectural differences between NVidia and AMD but XFX are generally lower cost OEM.  Higher DOA don't surprise me.  They have good warranties though.  My impression (via dead cards and sometimes illogical bioses) is they are the Kia motors of videocards.

[ztex]

But I am *more* than interested in acquiring a board filled with FPGAs (i.e. 5 daughterboards in the backplane?) - under the conditions that:

Maybe this is what you are searching for: https://bitcointalk.org/index.php?topic=49180.0

a) The kit is assembled to the point where the end-user (i.e. me) doesn't need to do any soldering more complicated than, say, splicing a custom connector to a PC standard PSU. I'm not an EE, not even an electronics hobbyist, and do NOT want to fuck up $1k with a clumsy soldering iron;

No soldering is required. A description about how a standard ATX PSU can be modified (without soldering ) for powering a rig can be found in the initial post of the topic mentioned above.

b) Getting the FPGAs mining away (pool or solo) is easy enough for a general-purpose software hacker and doesn't require EE knowledge. I mainly run Macs (because they're Unix with MS Office and work well) but all my mining rigs are Linux. I'd like to have my FPGA rig controlled by a Mac Mini or my old G4 Cube (CPU arch may cause problems if x86 libs are needed, though). I've only got 30 years coding experience but the lowest level code I know is C - unrolling loops and VHDL are *well* outside my skillset and I don't have time to learn;

The software (see http://www.ztex.de/btcminer) is ready-to-use and runs on Linux. Rigs can be controlled by a single instance using the cluster mode. Hot-plugging is supported too.

c) Apart from the peripheral software, everything is pre-loaded and coded. I am not familiar with FPGAs but know that the dev kit for the units talked about here costs a fortune. I won't be tuning the code and re-loading it onto a set of 5 FPGAs, so I don't want or need that cost, but I need it to run as soon as I plug in a cable and ping some control commands down the cable;

Bitstream (and Firmware) is compiled and ready-to-use. Firmware and Bitstream are uploaded by the software through USB. No JTAG programming cables or so are required.

d) The code loaded onto the FPGAs is *reasonably* efficient and not hugely sub-optimal. I don't want to spend a grand, and then find out in a couple of months about new bitstreams for the FPGAs I own... which would double my hashrate if I could re-program the things. I don't know how to do that, and I assume the SDK is needed too. From what I've read, this will not be a problem as the FOSS logic and all the proprietary optimisations aren't miles away from each other in speed?

The software typically achieves 190MH/s per XC6SLX150-3 FPGA.

d) ALL necessary cables are included - if they're custom then I'm happy to make them, but you HAVE to include the plugs / sockets because they may not be easily available to me in the UK (and if the connectors have 10+ pins then I'd prefer to pay for pre-made cables);

Only standard cables (which can be purchased in internet) are required.

e) You are happy to ship to the UK. I will assume trust once I've spoken to you via email so am happy to provide payment up-front so long I feel everything is legit. I won't waste your time.

Article location is Germany, i.e. unless you have not valid VATIN you have to pay 19% German VAT. (But if you import from outside the EU you also have to pay UK import VAT.)

[rph]

You know FPGA mining is becoming legit, when 2-3 vendors are trying to snipe customers from each others' threads.  

-rph

[fizzisist]

You know FPGA mining is becoming legit, when 2-3 vendors are trying to snipe customers from each others' threads. 

Haha, very true! Catfish, the truth is that all of the FPGA mining products you see here can be run by anyone who has managed to mine on a GPU. In fact, I think they are even easier to use (less complicated driver installs, overclocking, fan speeds, etc., and no need to even open up your tower to install it).

[rph]

Haha, very true! Catfish, the truth is that all of the FPGA mining products you see here can be run by anyone who has managed to mine on a GPU. In fact, I think they are even easier to use (less complicated driver installs, overclocking, fan speeds, etc., and no need to even open up your tower to install it).

Yeah, I'd agree with that. I just hope you guys don't sell so many that the difficulty becomes driven by FPGAs instead of GPUs.
Create some OPEC-style quotas or something..

-rph

[DeathAndTaxes]

Haha, very true! Catfish, the truth is that all of the FPGA mining products you see here can be run by anyone who has managed to mine on a GPU. In fact, I think they are even easier to use (less complicated driver installs, overclocking, fan speeds, etc., and no need to even open up your tower to install it).

Yeah, I'd agree with that. I just hope you guys don't sell so many that the difficulty becomes driven by FPGAs instead of GPUs.
Create some OPEC-style quotas or something..

-rph

Well given that FPGA have a long hardware payoff period I don't see FPGA putting much downward pressure on prices.

It will however put a floor on hashing power.  GPU are very electrical dependent.  At $3 per BTC @ current difficulty translates into roughly $0.15 per kWh on even the most efficient GPU.  Thus people who's electrical price is above the break even tend to quit and push hashing power & difficulty down.

FPGA however once bought are a sunk cost and have an electrical cost of <$0.50 per BTC meaning that they will continue to run likely no matter what.  The FPGA portion of hashing power already purchased (currently ~0%) will be "immune" to price changes.  What that means is as the FPGA portion grows the relationship between price/difficulty and hashing power will be come less linear. 

Even if prices spike I don't see a massive rush to buy FPGA but rather a slow continual rollout.  The long hardware payback period will make miners more cautious.  As an example when BTC prices hit $30 at 1.5M difficulty it became a no brainer to buy more GPU.  Even unsustianble as that would be.  The payback period was like 40 days.  If you mined for 40 days you could payoff a card.  FPGA however would still need a significant period of time to payoff the hardware so price spikes will have less influence on sales.


It will be an interesting dynamic to watch because I am sure 2012 will be the year of the FPGA.

[gmaxwell]

You know FPGA mining is becoming legit, when 2-3 vendors are trying to snipe customers from each others' threads.  

I prefer the kind of evidence where the margins get down to 20% over COGS.

[DeathAndTaxes]

So what is release date?  You may want to update title so people don't think the project is dead.

Got a quesiton about the backplane.  You say it has a 48W powersupply but also it has a SATA power connector?  By PSU do you mean it steps down the voltage from SATA power connector what is required for each board?

Any discounts if someone buys a backplane + 6 boards?
How heavy are the boards?  Your photo has the backplane "down" and the cards plugged into it.  Would there be an issue if the backplane was mounted vertically and the cards acting like "shelves" or would that put too much pressure on the connector.  Just getting some ideas on density and mounting.  I would want to have them enclosed in a case.

[eldentyrell]

So what is release date?

Sorry, I wasn't aware that the title could be altered!  I've updated it now.

Got a quesiton about the backplane.  You say it has a 48W powersupply

I've switched to a 72W supply.

but also it has a SATA power connector?  By PSU do you mean it steps down the voltage from SATA power connector what is required for each board?

Exactly.  Technically it is a "DC-to-DC point of load switching regulator."

Nice part is that the 72W supply can feed from either +5V or +12V (the old 48W supply could only use +12V).

Any discounts if someone buys a backplane + 6 boards?

At the moment I am neither taking orders nor announcing a ship date nor guaranteeing that either of these things will happen.  If you have immediate need for an FPGA mining device I suggest you look into the fine offerings by fpgaminer/ngzhang or ztex (or rph although I think he said he's not selling his).

How heavy are the boards?  Your photo has the backplane "down" and the cards plugged into it.  Would there be an issue if the backplane was mounted vertically and the cards acting like "shelves" or would that put too much pressure on the connector.

That works fine.  Those connectors are seriously heavy-duty stuff.  Unfortunately they're expensive too: even in qty50 I still had to pay $2.60 per board for each pair of connectors (male+female).  But there's almost no voltage drop across the gigantic pins and they can carry more current than I'll ever need.

[DeathAndTaxes]

Well I am not looking to buy until January.  I am disappointed you aren't looking to make a commercial run.  Still hopefully you make your personal run, learn some things and come back with a "round 2" offering.

When it comes to mining I always think big.  I replaced all my hodgepodge collection of GPU with nothing but 5970s because I like the density (2.2GH/rig without extenders or dual power supplies).  That kind of thinking let me get 10GH/s in my garage.  I like your board design because once again ... density.

10GH would be ~ 50 FPGA boards.  Now I have no intention on buying 50 boards all at once but I also like to plan for the "end game".  50 boards lying around and connected by a rats nets of USB cables doesn't seem appealing to me.  Maybe it is my times working in a datacenter or maybe it is just OCD but I like to see everything in their place.

Your design seems to have higher density and provide for more efficient racking.  One backplane w/ 6 cards ~= 1.2GH/s.  If you ever offered a larger backplane of say 10 cards powered by a single PCIe connector well that is even more interest.  Take a 4U server case put 2 backplanes and a tiny ITX motherboard in it.  A low end 500/600W power supply w/ 2 PCIE power connectors could power the entire thing.  20 cards or ~4GH/ in a single 4U case.   Power density would even be low enough to rack these things up in a standard datacenter rack.

Anyways even if you don't sell any in the near future I hope you keep on the project. If you make changes for "round 2" think about density.  It is the advantage you have over other designs, an advantage some would be willing to pay a slight premium for.  Open air rigs (either GPU or FPGA) are fine for now but the "end game" goal would be high hashing density in standardized server cases.  Nobody wants a garage or office full or whirling, noisy open air rigs they just happen to be the most efficient.  GPU likely will never work in standard case do to high thermal load but FPGA ... might.

[bulanula]

FPGA is much better for me than GPUs because less heat and noise right now but the price and performance leaves much to be desired.

Guess you can say the GPUs are old, inefficient, powerful gas guzzler motors while the FPGAs are new electric vehicles to keep the car analogy going

[DeathAndTaxes]

FPGA is much better for me than GPUs because less heat and noise right now but the price and performance leaves much to be desired.

Guess you can say the GPUs are old, inefficient, powerful gas guzzler motors while the FPGAs are new electric vehicles to keep the car analogy going

However they are getting close.  While GPU may be cheap there is a limit on how many can be powered by a single rig and high efficiency power supplies aren't cheap either.  A good price point for a GPU rig is $1 per MH/s.   FPGA are getting closer every day.

GPU rig $1 per MH & 2MH/W @ $0.10 per kWh.
1GH rig = $1000 hardware cost + $438 per year.  Total cost over 2 years = $2314.

FPGA Rig (22MH/W @ $0.10 per kWh)
1GH rig (@ $2.50 per MH) = $2500 hardware costs + $40 per year.  Total cost over 3 years = $2620.
1GH rig (@ $2.00 per MH) = $2000 hardware costs + $40 per year.  Total cost over 3 years = $2120.
1GH rig (@ $1.50 per MH) = $1500 hardware costs + $40 per year.  Total cost over 3 years = $1620.

Given FPGA massively lower operating costs if they even get close to GPU they are the smart place to deploy new hardware.

[ElectricMucus]

Why aren't there any pcie cards with only a few fpgas & some power converters on it?

That would be the most cost effective solution as modern fpgas have native pcie endnodes and pcie even has a jtag interface built in. All we need is drivers, a 2 layer pcb and a panel sheet to mount it.

[eldentyrell]

Why aren't there any pcie cards with only a few fpgas & some power converters on it?

Lack of demand.

Other than bitcoin, there are not many uses for large FPGAs outside of (1) development boards and (2) integration into a product of some kind (like a router) in which the user is not even aware that FPGAs are involved.  The second one is where Xilinx's large-device profits come from.

If you're buying a dev board, you're either an academic (in which case Xilinx cuts you a special deal) or you don't mind paying $4,000+ for a card with all sorts of doo-dads you'll never use since odds are the development costs of the product you're working on make expenses like this irrelevant.  That's why with each generation of chips you see Xilinx (or one of its partners) produce some uber-board with everything-and-the-kitchen-sink on it.  FWIW, many of these "kitchen sink" boards have PCIe interfaces.

[ngzhang]

Why aren't there any pcie cards with only a few fpgas & some power converters on it?

That would be the most cost effective solution as modern fpgas have native pcie endnodes and pcie even has a jtag interface built in. All we need is drivers, a 2 layer pcb and a panel sheet to mount it.

Why use PCIE interfaces? USB1.1 is much better.

[DeathAndTaxes]

Why aren't there any pcie cards with only a few fpgas & some power converters on it?

That would be the most cost effective solution as modern fpgas have native pcie endnodes and pcie even has a jtag interface built in. All we need is drivers, a 2 layer pcb and a panel sheet to mount it.

Why use PCIE interfaces? USB1.1 is much better.

I am thinking scalability and density.  Evetually if Bitcoin grows and flourishes, mining will move beyond hobbyist and garages full of open rigs full of noisy cards into high density datacenter capable designs.  Getting large number of GPU in a rack mount server is unviable due to the thermal load.  FPGA make that possible someday.  A PCIe board can supply power and data over single connector which makes deployment easier.  More importantly it provides a way to securely mount multiple FPGA using existing standards (off the shelf motherboards, rackmount chassis, ATX power supplies, etc). 

I would love someday to be able to put a FPGA array in a co-location datacenter to reduce risk of loss due to theft, power, fire, damage. A full length board would be able to mount maybe 5 FPGA for half-height board and maybe 10 for full height board.  That creates some interesting datacenter quality arrays.  A 2U server could mount 4 boards, 20 FPGAs (or more).  That is ~4GH/s on 300W in a 2U space.  A standard datacenter rack could hold a 80GH/s of hashing power, run on a single 30A 208V power connection and make things like remote power control, KVM over IP, and enterprise grade redundant power supplies more economical.

[bulanula]

Why aren't there any pcie cards with only a few fpgas & some power converters on it?

That would be the most cost effective solution as modern fpgas have native pcie endnodes and pcie even has a jtag interface built in. All we need is drivers, a 2 layer pcb and a panel sheet to mount it.

Why use PCIE interfaces? USB1.1 is much better.

I am thinking scalability and density.  Evetually Bitcoin will move beyond hobbyist and open boards into high density datacenter designs.  Getting large number of GPU in a rack mount server is simply impossible due to the thermal load.  FPGA make that possible someday. I see that as the endgame for FPGA.  A PCIe board can supply power and data over single connector.  It also make a convinent way to mount multiple FPGA inside a standardized chasis.  I would love someday to put a FPGA array in a co-location datacenter to reduce risk of loss due to theft, power, fire, damage.  

I full length board would be able to mount maybe 5 FPGA for half-height board and maybe 10 for full height board.  That creates some interesting datacenter quality arrays.  2U server could mount 4 boards or 20 FPGA total for ~4GH/s using maybe 300W for entire system  (at the wall).  A standard datacenter rack could hold a 80GH and run on a single 30A 208V power connection.  The higher density would make things like remote power control and KVM over IP economical.

Too bad the demand is too low now. I think BFL labs is scam too. I mean why go through all that development when the price of BTC can crash any day now and people will stop buying mining equipment etc. Even in other industries FPGA is almost never heard of. I never heard about FPGAs until Bitcoin etc.

[aTg]

I am thinking scalability and density.  Evetually Bitcoin will move beyond hobbyist and open boards into high density datacenter designs.  Getting large number of GPU in a rack mount server is simply impossible due to the thermal load.  FPGA make that possible someday. I see that as the endgame for FPGA.  A PCIe board can supply power and data over single connector.  It also make a convinent way to mount multiple FPGA inside a standardized chasis.  I would love someday to put a FPGA array in a co-location datacenter to reduce risk of loss due to theft, power, fire, damage.  

I was thinking exactly that, but we could not start from here with that standard design for a rack?
I think that having many small modules with a single FPGA is not efficient for spending on individual fans and especially because a single USB controller could handle an entire plate of FPGA's so each module in the rack may be connected via USB to a hub and a computer within the same cabinet.

[DeathAndTaxes]

I was thinking exactly that, but we could not start from here with that standard design for a rack?
I think that having many small modules with a single FPGA is not efficient for spending on individual fans and especially because a single USB controller could handle an entire plate of FPGA's so each module in the rack may be connected via USB to a hub and a computer within the same cabinet.

Agreed but PCIe "solves" 3 solutions

1) power distribution
2) data connectivity
3) standardized mounting
4) server sized cooling not individual board cooling

Sure you could have larger boards, and figure out a way to rig usb cables to a hub to the host, run custom power lines to each of them, and then figure out some non-standard method to securely mount and cool them  However using PCIe allows you to leverage existing technology like chassis with redundant midplane cooling, backplanes for securely mounting cards, ATX motherboards for connectivity and power.  I don't think we will see PCIe solutions anytime soon but on the other hand I can't imagine if Bitcoin is around in 5 years that the "solution" is a bunch of usb boards jury rigged inside a case connected to usb hub.

For example take a look at this "industrial chassis"
http://www.adlinktech.com/PD/marketing/Datasheet/RK-440/RK-440_Datasheet_1.pdf

Notice the midplane fans designed to cool expansion cards and the 18 expansion slots.  It uses a "single board computer" where the "motherboard" is actually mounted perpendicular to a backplane just like any other expansion card.  This is the kind of setup that is used for other "industrial" servers like cable video mulxiplexing, high speed network switching, digital signal processing, etc. 

[ztex]

I don't think we will see PCIe solutions anytime soon but on the other hand I can't imagine if Bitcoin is around in 5 years that the "solution" is a bunch of usb boards jury rigged inside a case connected to usb hub.

Development costs are much higher (driver development, ...). Due to the small amounts sold this results in significant higher prices of such boards.

However future solutions will look like, unless you do not want to invest $100000s it will either be ugly or cheap.

[Dexter770221]

FPGA chip that can give resonable MH/$ (1?) cost at least 150$. If you want to put 6 of them to one card that gives 900$ for chips only. PCB and other necessary parts that would be 300$ more. And 300$ for manufacture cost and spedition. Thats 1500$ per card that can only mine and achive ~1.2GH/s (using ~50W of power). Thats quiet a big price... When I've started my adventure with bitcoin I spend 1000$ for PC that can produce 800MH/s and that wasn't easy decision... I've done that becuse I'm using PC also for work (writing programs). My child needs new pair of boots, so I will never spend 1500$ for some card that can be worthless in few months... Propably most of bitminers are in familiar situation... Thats why there's no such card, to small demand. Give me 2k$ and i will design such a card and make one prototype...

[bulanula]

I was thinking exactly that, but we could not start from here with that standard design for a rack?
I think that having many small modules with a single FPGA is not efficient for spending on individual fans and especially because a single USB controller could handle an entire plate of FPGA's so each module in the rack may be connected via USB to a hub and a computer within the same cabinet.

Agreed but PCIe "solves" 3 solutions

1) power distribution
2) data connectivity
3) standardized mounting
4) server sized cooling not individual board cooling

Sure you could have larger boards, and figure out a way to rig usb cables to a hub to the host, run custom power lines to each of them, and then figure out some non-standard method to securely mount and cool them  However using PCIe allows you to leverage existing technology like chassis with redundant midplane cooling, backplanes for securely mounting cards, ATX motherboards for connectivity and power.  I don't think we will see PCIe solutions anytime soon but on the other hand I can't imagine if Bitcoin is around in 5 years that the "solution" is a bunch of usb boards jury rigged inside a case connected to usb hub.

For example take a look at this "industrial chassis"
http://www.adlinktech.com/PD/marketing/Datasheet/RK-440/RK-440_Datasheet_1.pdf

Notice the midplane fans designed to cool expansion cards and the 18 expansion slots.  It uses a "single board computer" where the "motherboard" is actually mounted perpendicular to a backplane just like any other expansion card.  This is the kind of setup that is used for other "industrial" servers like cable video mulxiplexing, high speed network switching, digital signal processing, etc. 

Hey D&T :

Since you seem very knowledgeable on these damn backplates I have one question for you. Do you think they can be used to mine with them etc. ?

I researched this and I think it is quite nice solution but not that cheap unfortunately. Eg will the costs be insane ? Power distribution in backplane ? Bandwidth in backplane etc. ? Thanks !

[aTg]

Something like that?

http://jchblue.blogspot.com/2009/08/pico-computing-fpga-cluster.html

16 Xilinx Spartan XC3S5000 FPGAs

[DeathAndTaxes]

Nobody seriously looking to buy FPGA is looking to buy a single board and leave it at that.  Just like nobody runs a hashing farm today with a single GPU.  I mean a single board running @ 200MH will generate roughly $0.60 per day.  Hardly something to get excited over. It will take tens of thousands of those FPGA to displace GPU as the dominant technology.  That won't be done by fifty thousand people buying a single board it will be done by couple hundred people by hundreds of chips and that is easier done with more scalable solutions.

Regardless if you have 1 FPGA or a dozen of them you still need a host, you still need a power supply, you still need to monitor it (time value).  It is a lot easier to justify one's time if generating $60 a day than $0.60 per day.  Like I said I understand WHY FPGA is at the stage it is today.  You have to start somewhere and Serial/USB single chip FPGA boards are an easier place to start.  However just because we start here doesn't mean we will end up here.

[DeathAndTaxes]

Something like that?

http://jchblue.blogspot.com/2009/08/pico-computing-fpga-cluster.html

16 Xilinx Spartan XC3S5000 FPGAs

I think something more like this ...


Kinda hard to tell from the photo but the FPGA are on daughter boards.  There are 6 expansion slots on each side (2 are populated).  Now granted they are high end Virtex FPGA, onboard memory, and PCIe 16x connector so not economical for mining.  Still the same kind of concept could be done for lower end chips. A single large PCIe board with room for up to 12 daughter cards.  You can buy 1 board and 1 to 12 FPGA depending on your budget.

Look I hope the FGPA authors/designers don't take this as bashing.  What they have accomplished is amazing. It has really brought the pricepoint for FPGA from "pie in the sky" to expensive but viable.  I am just saying long term scalability will matter.  A board like this would allow someone to go from 1 to 48 FPGA in a single server (200 MH/s to 9600 MH/s).  Everything other than the card will be standard data center stuff.

[Dexter770221]

Nice looking monster Price is propably somwhere on stratosphere And performance rather poor. That single Spartan 3S5000 maybe can reach 50MH/s (*16=800MH/s).

[DeathAndTaxes]

I researched this and I think it is quite nice solution but not that cheap unfortunately. Eg will the costs be insane ? Power distribution in backplane ? Bandwidth in backplane etc. ? Thanks !

Yes cost will be high.  Backplanes tend to run $300 to $1000+.  The single board computer (mother board equivalent) runs another $200.  The way GPU economics work there really is no advantage to putting more than 6 or so GPU per board.  The main advantage of a backplane would be ability to put it all in a chassis but the thermal load of 10+ GPU makes that totally impossible so there really is no point.

[fizzisist]

Kinda hard to tell from the photo but the FPGA are on daughter boards.  There are 6 expansion slots on each side (2 are populated).  Now granted they are high end Virtex FPGA, onboard memory, and PCIe 16x connector so not economical for mining.  Still the same kind of concept could be done for lower end chips. A single large PCIe board with room for up to 12 daughter cards.  You can buy 1 board and 1 to 12 FPGA depending on your budget.

I actually looked into that board. They sell a similar one with Spartan 6 LX150s on it actually. The daughter boards have 2 FPGAs each. The price per daughter board? Around $3000!

Here's the pic:


The backplane card itself is about $1,500. That means the fully loaded card costs about $20,000. That's 12 FPGAs, or 1.5-2 GH/s or so. All told, $10 / MH/s! If you only bought one daughterboard, it would be $4500 for only about 300 MH/s!!

Finding this out was the final straw that drove me to start working on an FPGA board specifically for bitcoin mining. It just seemed insane that they could get away with charging so much. I knew that it could be done for less. Unfortunately, part of the cost savings was eliminating the PCIe interface.

For us, USB is a natural decision, but I'm not opposed to other designs. If you really want to build a PCIe based system like that, and think others would be interested in it, I'd be happy to work with you to design it together. Let's talk!

[aTg]

For a PCI expansion chassis with 13 connections over 2mil dollars ... we have a problem, we would have to design ourselves all hardware including PCI.

Another solution I can think of is to design an FPGA circuit that is powered by USB, one PC card for pci usb expansion could feed many circuits with multiple FPGAs



This is a quick sketch

[bulanula]

For a PCI expansion chassis with 13 connections over 2mil dollars ... we have a problem, we would have to design ourselves all hardware including PCI.

Another solution I can think of is to design an FPGA circuit that is powered by USB, one PC card for pci usb expansion could feed many circuits with multiple FPGAs



This is a quick sketch

Don't quite understand what you are saying. Why not just use USB FPGAs and USB hubs etc. I though bitcoin mining is supposed to be "ghetto".

[ElectricMucus]

Something like that?

http://jchblue.blogspot.com/2009/08/pico-computing-fpga-cluster.html

16 Xilinx Spartan XC3S5000 FPGAs

These with  16 Spartan-6 LX150s would have to come close to 3 GH/s, a miners wet dream if they were affordable. But most likely they have a huge premium on the chips themselves.

[ElectricMucus]

Why aren't there any pcie cards with only a few fpgas & some power converters on it?

That would be the most cost effective solution as modern fpgas have native pcie endnodes and pcie even has a jtag interface built in. All we need is drivers, a 2 layer pcb and a panel sheet to mount it.

Why use PCIE interfaces? USB1.1 is much better.

How better?
With pcie you could spare:
USB chip, jtag connector, power connector, programming cable
With USB you spare:
implementing the pcie endnote according to free xilinx instructions, a level shifter ic and a few CLBs

[DeathAndTaxes]

implementing the pcie endnote according to free xilinx instructions, a level shifter ic and a few CLBs

Yes, on the *T model chips that have an on board pci-e interface and which cost something like 15% more.

Then you're stuck attaching the damn things to a motherboard... and confined to the computer cases which are available or can be constructed.

Far better to have a USB (or ethernet, or RS422, or anything _not_ pcie) interface — then you can build an enclosure which meets your density and cooling needs optimally— e.g. nice linear airflow because your system is a single plane, and you can service hundreds of miner chips off a pair of standard rackmount servers.

I doubt you are going to build an enclosure including all data transfer, cooling, and power cheaper than bulk purchased 2U & 4U rackmount chassis.

Density is a non-issue if using PCIe cards.  You can radiate about 150W from an expansion slot using chassis mounted fans and passive heat sink.  That is now NVidia cools Tesla cards.  FPGA are so power efficient that 150W is something like 3GH/s.  

So there is no need to build an enclosure to "meet your needs".  The enclosure that meets your needs is a standard 4U server.  Put 16 FGPA on a single PCIe card and you got 3.2GH/s per expansion slot.  4 cards per server works out to "only" 12.8GH per 4U server.  Standard data center rack could hold 128GH.  If for some reason you needed more capacity they make Single Board Systems with 12+ PCIe expansion slots in standard rackmount chassis.  So you got standardized and commoditized expansion, cooling and power distribution options.

Granted USB is easier to get working but eventually cost effective PCIe cards will come along.  PCIe isn't that expensive.  



If a $14 NIC can implement PCIe eventually someone will figure out how to cost effectively add PCIe support to $2000 worth of FPGAs.

[DeathAndTaxes]

Have you priced out any of those backplanes?

Have you priced out the non-PCIe boards.  Hell the dual Spartan-6 150 module (just the FPGA and a couple MB of RAM) is $4000.  By that logic nobody can make a mining board for less than $2K. 

If a NIC can implement a PCIe interface for the same price as a USB interface then obviously there isn't anything magical that makes FPGA add $1500 to the cost on the board.  The prices a high because the margins are insane in semi-custom FPGA land.  The margins are equally high on that companies USB, compact flash, and serial modules too.  That doesn't tell us anything about the actual cost to implement. 

[rph]

PCI-e is just a pain in the ass for mining. You need gold finger controlled impedance PCBs, a
flash PROM to configure the FPGA before the PC BIOS scans the bus, much more complex
interface logic in the FPGA, more complex SW, and possibly a $$$$ high speed scope if
something goes wrong, just to provide insane amounts of bandwidth you don't even need.

Plus, putting 16 $150 BGAs onto a single giant PCB is a bad idea. 1 out of every 20-30 boards will have
a critical defect under 1 of the FPGAs, costing you either $2400 to scrap the board, or a bunch of time and
money to try to remove/reball/repair it. There's a good manufacturability reason to use only 1-2 FPGAs per PCB -
if something goes very wrong you're out $300, not $2000.

-rph

[ElectricMucus]

What the hell, with a high speed CPLD you can probably bitbang pcie, last time I checked those cost around $5.
Some of those can interface with almost and logic level & impendace.

The thing what concerns me is: Why use a PC anyway? We could possibly implement the mining software on a cheap ARM, pair it with a ethernet MAC ic and be done with it.

[DeathAndTaxes]

What the hell, with a high speed CPLD you can probably bitbang pcie, last time I checked those cost around $5.
Some of those can interface with almost and logic level & impendace.

The thing what concerns me is: Why use a PC anyway? We could possibly implement the mining software on a cheap ARM, pair it with a ethernet MAC ic and be done with it.

Scalability.  If you wanted to build a 20G/s cluster are you going to have 100 cases, 100 boards, 100 ARM processors, 100 ethernet ports, 100 wallwart power-supplies?

[ElectricMucus]

Just try to jam as many fpgas on the board as possible, I know the whole fault rate issue as described by rph... Just make one faulty board cost less to repair than a pc and you're done.

However: 1HE 48port switch, 47 1HE units * 64 Spartan-6 LX150 * 180MH = 541 GH/Rack
Ok costs would probably kill it the fpgas alone would cost almost half a million dollar  

[ngzhang]

PCI-e is just a pain in the ass for mining. You need gold finger controlled impedance PCBs, a
flash PROM to configure the FPGA before the PC BIOS scans the bus, much more complex
interface logic in the FPGA, more complex SW, and possibly a $$$$ high speed scope if
something goes wrong, just to provide insane amounts of bandwidth you don't even need.

Plus, putting 16 $150 BGAs onto a single giant PCB is a bad idea. 1 out of every 20-30 boards will have
a critical defect under 1 of the FPGAs, costing you either $2400 to scrap the board, or a bunch of time and
money to try to remove/reball/repair it. There's a good manufacturability reason to use only 1-2 FPGAs per PCB -
if something goes very wrong you're out $300, not $2000.

-rph

totally  agree with you.

What the hell, with a high speed CPLD you can probably bitbang pcie, last time I checked those cost around $5.
Some of those can interface with almost and logic level & impendace.

The thing what concerns me is: Why use a PC anyway? We could possibly implement the mining software on a cheap ARM, pair it with a ethernet MAC ic and be done with it.

on my design, the upper computer only need send and receive mining datas, so a hacked router is enough.

What the hell, with a high speed CPLD you can probably bitbang pcie, last time I checked those cost around $5.
Some of those can interface with almost and logic level & impendace.

The thing what concerns me is: Why use a PC anyway? We could possibly implement the mining software on a cheap ARM, pair it with a ethernet MAC ic and be done with it.

Scalability.  If you wanted to build a 20G/s cluster are you going to have 100 cases, 100 boards, 100 ARM processors, 100 ethernet ports, 100 wallwart power-supplies?

no case, 100 boards, 1 arm processor, 1 Ethernet ports, 2~3KW  12V power-supplies is ok.

[bulanula]

WOW FPGA really is picking some steam. I think ngzhang took it to the next level with his really neat professional board. Now only if the cost was reasonable

[aTg]

PCI-e is just a pain in the ass for mining. You need gold finger controlled impedance PCBs, a
flash PROM to configure the FPGA before the PC BIOS scans the bus, much more complex
interface logic in the FPGA, more complex SW, and possibly a $$$$ high speed scope if
something goes wrong, just to provide insane amounts of bandwidth you don't even need.

Plus, putting 16 $150 BGAs onto a single giant PCB is a bad idea. 1 out of every 20-30 boards will have
a critical defect under 1 of the FPGAs, costing you either $2400 to scrap the board, or a bunch of time and
money to try to remove/reball/repair it. There's a good manufacturability reason to use only 1-2 FPGAs per PCB -
if something goes very wrong you're out $300, not $2000.

-rph

That oblem is solved using a socket on the motherboard or something like what you have invented in your design.

[Albert]

I must say, months ago when I first stumbled upon the bitcoin, I lost hope in ever utilizing it as a online payment alternative because of GPU+Power costs, and the fact that the value dropped greatly within 2 months (Back when I was concerned with profiting from it, alone).

I'd be able to afford these up-and-coming FPGA boards, and would be willing to tag along with a few pools. I am devoted again thanks to these recent advancements.

What'll they think up next? 

[DeepBit]

I must say, months ago when I first stumbled upon the bitcoin, I lost hope in ever utilizing it as a online payment alternative because of GPU+Power costs, and the fact that the value dropped greatly within 2 months (Back when I was concerned with profiting from it, alone).

May be you misunderstood how the bitcoin works ?
You don't need to mine bitcoins for using it as online payment system. You just earn or buy coins when you need them.

[Albert]

I'm well aware of how I can buy bitcoins. However, my intention is not to simply buy into them; after all, market trends show that to be somewhat negligible a route for someone expecting to profit from this. No, my only intention is to add to the market and trade as much as I can on the coin alone.

Course, I could buy just a coin or two, but I'm just going to get into mining. And besides that, I'm not very confident in my trade skills.

[n0ne]

Updates on the project?

[ZodiacDragon84]

Interesting

[punningclan]

Thats a very nice looking board at a competitive price, subscribing!

[mokahless]

How are you boards coming along? Any more details on your hash rates/problems you encountered/have you given any more consideration to selling?

[eldentyrell]

How are you boards coming along?

I've been mining on them for several months now.

have you given any more consideration to selling?

No, I don't really want to be in the board-making business.

[IVIasterZox]

How does it look today?

[vs3]

Just curious - is this project still alive?

[eldentyrell]

Just curious - is this project still alive?

I'm not selling them, it is simply not profitable.

I'm running quite a large number for myself.

[Slave2school]

That's a shame

[Skriglitz]

Just curious - is this project still alive?

I'm not selling them, it is simply not profitable.

I'm running quite a large number for myself.

Such a shame indeed..  However is it possible to get a copy of the schematics and pcb files so that I can try myself?

[razorfishsl]

Just curious - is this project still alive?

I'm not selling them, it is simply not profitable.

I'm running quite a large number for myself.

Such a shame indeed..  However is it possible to get a copy of the schematics and pcb files so that I can try myself?

It is not worth it....... At the current hash rate and difficulty , you have about 1 month before you loose money....

Run the costings
http://www.bitcoinx.com/profit/

But.... if it is a research project you can check out my work here...
http://hardcoreforensics.com
At least at the end of it you will have a scrap board that has a significant amount of memory, PSU's and a PCI interface...

[neslekkim]

But.... if it is a research project you can check out my work here...
http://hardcoreforensics.com
At least at the end of it you will have a scrap board that has a significant amount of memory, PSU's and a PCI interface...

Did you ever look at the Pano Logic G2 modules? http://www.ebay.com/itm/200880792833
The G2 version have Spartan6 LX150 in them, would have been cool to do something about it, I have one, and will read your project to see if I can learn something.

Hackaday wrote about them earlier this year: http://hackaday.com/2013/01/11/ask-hackaday-we-might-have-some-fpgas-to-hack/