ZTEX USB-FPGA Modules 1.15x and 1.15y: 215 and 860 MH/s FPGA Boards

[DeathAndTaxes]

Do you need a special type of USB port?  You say it uses 8.5w, and a USB port provides 500mA @ 5v = 2.5w.  While some USB ports can provide more amperage than that, the standard only requires 500mA...

It uses a power adapter (visible on PCB below USB on left hand side).  
My understanding (ztek can correct me) is that the USB port is just used for communication.

[m3sSh3aD]

Do you need a special type of USB port?  You say it uses 8.5w, and a USB port provides 500mA @ 5v = 2.5w.  While some USB ports can provide more amperage than that, the standard only requires 500mA...

It uses a power adapter (visible on PCB below USB on left hand side).  
My understanding (ztek can correct me) is that the USB port is just used for communication.

Correct

[DeathAndTaxes]

About the power...
Ztek why the decision to use Center pin DC instead of just using 12V Molex connector.

I could see rigging an adapter to install them in 3.5" HDD bay (some cases have 12+ 3.5" bays).  It would be conveinent to power multiple boards via computer power supply MOLEX connectors.  I am thinking of ultra low end CPU integrated MB running a lightweight linux distro (lubuntu, puppy linux, etc) as the host for multiple boards.

Maybe I could rig up some Molex to DC center pin adapter.  See any problem with that?

[ztex]

And the community is requesting that you don't have to request.

yeah, and BTC prices updated once per month ...

I'm sorry, but currently I'm to busy to implement an automatic solution in the shop. (I spent more time than planned for the BTCMiner software)

[ztex]

Power: CON5 (see http://www.ztex.de/usb-fpga-1/usb-fpga-1.15x.e.html#con5) is used for external power supply.

USB connector is only used for communication.

About the power...
Ztek why the decision to use Center pin DC instead of just using 12V Molex connector.

I could see rigging an adapter to install them in 3.5" HDD bay (some cases have 12+ 3.5" bays).  It would be conveinent to power multiple boards via computer power supply MOLEX connectors.  I am thinking of ultra low end CPU integrated MB running a lightweight linux distro (lubuntu, puppy linux, etc) as the host for multiple boards.

Maybe I could rig up some Molex to DC center pin adapter.  See any problem with that?

The disadvantage of HDD or SATA connectors is that the amount of boards which can be connected to one ATX PSU is limited to 20 or so. Y adapter cables will not help much since each cable from the PSU can only supply up 5 to 10 boards (3.5 to 7 A), depending on the cross-sectional area of the cable. You also have to use the graphic card cables and/or the 12V lines intended for the CPU.

As simple solution I recommend modular PSU's, i.e. the ones with plugable cables. Cut of the connectors from these cables and connect them to this http://search.digikey.com/scripts/DkSearch/dksus.dll?&KeyWords=CP-2189-ND using crimp tools.

[allten]

When you have an adapter where you can stack multiple FPGAs on top (or next to) of each other using the same communication board - then I'm ready to buy more than one.

I don't think you need any adapter however you can't just stack FPGA you can only stack boards.

The board has 4 screw cut outs.  Using standoffs (like what you mount motherboard with but longer) longer than the heatsink you could build a "tower" of these.  Might be better to turn them on their sides though and use a different heat sink which blows air across the heatsink not "up".

If what you were talking about is a board that had like 4 sockets that you could buy & plug in multiple FPGA that is interesting idea however that makes the cost of 1 FPGA higher (due to unused overhead) which likely means you need much bigger market to support that.  

I was referring to both ideas. I really like idea of stacking them on top of each other with small heat-sinks sandwiched in that you could have air blow through.
Here's an example:

http://www.microchip.com/stellent/idcplg?IdcService=SS_GET_PAGE&nodeId=1406&dDocName=en535770
Notice the male connector surrounding the IC. What if you had that on an FPGA board with the ability to stack multiple ICs?

Each IC would be soldered to one of these for stacking
https://bitcointalk.org/index.php?topic=44891.0

Maybe there are design limits that would make this unfeasible, but it seams you should be able to stack at least 2 additional ones with a board design that supports it. Unless someone can convince me that it is impossible, I will be holding out on buying multiple of these. I really do like the featured product on the OP; it really shows that we are evolving fast which makes me very excited.

Edited for typos

[SgtSpike]

Ah, my bad.  I didn't see a power connector on the board... must have skimmed the pictures too quickly. 

[amazingrando]

Edited: Original calculations overstated BTC/day for both FPGA and GPU.  Revised calculations show that FPGA makes sense at a much lower power cost.

It seems that the viability of this product rests completely on electricity pricing.  If you compare FPGA to GPU with the same amount of money invested it appears (using round numbers) that somewhere in the range of 12 8 cents per kWh is where either solution is roughly equal in profit.  If your power costs more than that, you're better off with an FPGA.  If you can get cheaper power, go for GPUs.

As an example, let's say you invest $2000 into mining.  With that, let's assume you could either buy 5x ZTEX FPGA modules or two GPU mining rigs with 8x GPUs.

FPGA
Investment: $2000
FPGA's: 5
Total Mh/s: 950
Monthly power cost at $0.08 for 45W: ~$3
BTC Payout: 0.6508 BTC/Day or ~20 btc/month
USD Payout at $2.70: $54
Monthly net revenue from $2000 investment (BTC minus power): $51

GPU
Investment: $2000
4-GPU rigs: 2 (each having 4 x Radeon 5850 GPUs producing 333 Mh/s each)
Total Mh/s: 2664
Monthly power cost at $0.08 1700W: ~$99
BTC Payout: 1.825 BTC/Day or ~55.5 btc/month
USD Payout at $2.70: $150
Monthly net revenue from $2000 investment: $51

Of course this comes with the standard caveats that BTC prices and difficulty will change and that you might be able to overclock GPUs and get even more Mh/s, etc. etc..  Also, if AMD can deliver on huge cuts in power consumption with the 7xxx series GPUs, the power indifference line will increase.

But, if you're considering investing in mining equipment and are looking at this FPGA or GPUs, your electricity rates may have already made the decision for you.

[ztex]

It seems that the viability of this product rests completely on electricity pricing.  If you compare FPGA to GPU with the same amount of money invested it appears (using round numbers) that somewhere around 12 cents/kwh is the indifference line between these two.  If your power costs more than that, you're better off with an FPGA.  If you can get cheaper power, go for GPUs.

As an example, let's say you invest $2000 into mining.  With that, let's assume you could either buy 5x ZTEX FPGA modules or two GPU mining rigs with 8x GPUs.

FPGA
Investment: $2000
FPGA's: 5
Total Mh/s: 950
Monthly power cost at $0.12 for 45W: ~$4
BTC Payout: 0.9762 BTC/Day or ~30 btc/month
USD Payout at $2.70: $81
Monthly net revenue from $2000 investment: $77

GPU
Investment: $2000
4-GPU rigs: 2 (each having 4 x Radeon 5850 GPUs producing 333 Mh/s each)
Total Mh/s: 2664
Monthly power cost at $0.12 1700W: ~$150
BTC Payout: 2.7375 BTC/Day or ~83.25 btc/month
USD Payout at $2.70: $225
Monthly net revenue from $2000 investment: $75

Of course this comes with the standard caveats that BTC prices and difficulty will change and that you might be able to overclock GPUs and get even more Mh/s, etc. etc..  Also, if AMD can deliver on huge cuts in power consumption with the 7xxx series GPUs, the power indifference line will increase.

But, if you're considering investing in mining equipment and are looking at this FPGA or GPUs, your electricity rates may have already made the decision for you.

Thanks for your calculations. But don't forget, for larger amounts (group order) you get higher volume discounts. See the initial message for bulk prices.

[ztex]

Maybe there are design limits that would make this unfeasible, but it seams you should be able to stack at least 2 additional ones with a board design that supports it. Unless someone can convince me that it is impossible, I will be holding out on buying multiple of these. I really do like the featured product on the OP; it really shows that we are evolving fast which makes me very excited.

Stacking these FPGA boards is not feasible due to several reasons

• Stackable pin headers are expensive and either have to be soldered by hand or have to be pressed in (both is expensive)
• At high currents you get problems with the contact resistance
• The space between the board is usually to small for a proper cooling
• Last but not least: At least all clock signals would have to be terminated (in order to avoid that reflected signals are misinterpreted as clocks). But termination is impossible since such a stack is everything but impedance controlled.

[eldentyrell]

Nice job!

Glad to see you added the northbridge sink mounting holes.  But you might want to consider putting in two sets of holes at different distances like I did -- there isn't an official standard for how far apart they're supposed to be, and the distance can vary by as much as 6mm -- enough that any single pair of holes will rule out a significant fraction of the heatsinks out there due to being either too close or too far.

Why the electrolytic capacitor?  I see you've got at least one polymer (i.e. does-not-explode-and-leak) cap on there... curious why you didn't use that for both of them.  I've switched to all polymer+ceramic.

Also curious: is there a reason why you aren't using the FGG484 package?  It's slightly cheaper than all the others (not sure which one yours is, but it's not the FGG484).

[ztex]

Why the electrolytic capacitor?  I see you've got at least one polymer (i.e. does-not-explode-and-leak) cap on there... curious why you didn't use that for both of them. 

Polymer capacitors are electrolytic capacitors.

I've switched to all polymer+ceramic.

... and to THT bypass capacitors

is there a reason why you aren't using the FGG484 package?

Yes,  the USB-FPGA Module 1.15x has to be firmware and bitstream compatible to USB-FPGA Module 1.15d.

[sadpandatech]

yours is 190 @ $460($2.50 m/hash) and there duel is 250 @ $580($2.35 m/hash).

There are volume discounts, see the initial post. If you order 5, the price is $430 or $2.26 per MH/s, the 25 units price is $1.94 per MH/s. If you search for an assembler and build an 50 GH/s rig, the price is about $1.02 per MH/s.

Even power usuage is in the same area

17W @ 250 MH/s: 14.7 MH/s per W
8.5W @ 190 MH/s: 22.3 MH/s per W

Thea is a difference of 50%.

ah yes, so it is The duel is less efficent but when you talking 50% on the amount of watts were talking its neither here nor there compared to other options (VGA etc). Great little devices all the same

  I don't know about that. ;p  At 50GHs that amounts to 1164 Watts less in electricty costs on this Ztex board. If electricty costs are at all an issue, this will add up pretty quick. ~28KW per day. Or $1,226 per year @ .12 KWH.  Yea, maybe not a big deal unless you're approaching it from a business perspective.

  Awesome board, Ztex!  Thanks for making it available to the community.

  

[Grinder]

FPGA
Investment: $2000
FPGA's: 5
Total Mh/s: 950
Monthly power cost at $0.12 for 45W: ~$4
BTC Payout: 0.9762 BTC/Day or ~30 btc/month
USD Payout at $2.70: $81
Monthly net revenue from $2000 investment: $77

Your BTC/month is almost 50% higher than the real earnings at these difficulty levels. They'll fall at the next change, but not enough to earn 30/month at 950 Mh/s. It seems to me that it won't fall this much unless the price falls even more.

[sadpandatech]

FPGA
Investment: $2000
FPGA's: 5
Total Mh/s: 950
Monthly power cost at $0.12 for 45W: ~$4
BTC Payout: 0.9762 BTC/Day or ~30 btc/month
USD Payout at $2.70: $81
Monthly net revenue from $2000 investment: $77

Your BTC/month is almost 50% higher than the real earnings at these difficulty levels. They'll fall at the next change, but not enough to earn 30/month at 950 Mh/s. It seems to me that it won't fall this much unless the price falls even more.

  You're right, its .65 BTC per day on the FPGA setup and 1.83 BTC per day on the GPU setup in the example at current difficulty. Maybe his calc uses a pool hopping formula to get the numbers?

[DeathAndTaxes]

50% would be high even for pool hopping.  Real world (especially today w/ pools putting up countermeasures and less pools being hoppable at all) hopping "only" nets you 25%-35%.

[amazingrando]

FPGA
Investment: $2000
FPGA's: 5
Total Mh/s: 950
Monthly power cost at $0.12 for 45W: ~$4
BTC Payout: 0.9762 BTC/Day or ~30 btc/month
USD Payout at $2.70: $81
Monthly net revenue from $2000 investment: $77

Your BTC/month is almost 50% higher than the real earnings at these difficulty levels. They'll fall at the next change, but not enough to earn 30/month at 950 Mh/s. It seems to me that it won't fall this much unless the price falls even more.

You're right.     Sorry about that.  I pulled the wrong numbers so the output was too high for both FPGA and GPU.  I have updated the calculations, which now shows that FPGA and GPUs are roughly equivalent at 8 cents/kWh, everything else being equal.

[gmaxwell]

You're right.     Sorry about that.  I pulled the wrong numbers so the output was too high for both FPGA and GPU.  I have updated the calculations, which now shows that FPGA and GPUs are roughly equivalent at 8 cents/kWh, everything else being equal.

People should run their own numbers in any case— your $/MH on the GPU farm is higher than my actuals by about 25% but people build with different degrees of efficiency.

You're certainly making the point that FPGAs are now in the realm of competitive for new deployments, even for buildouts with fairly short term planning horizons. Ztex's modules are also attractive because they've clearly been designed in a manner which would facilitate re-purposing them for things other than mining, which is a reason they should retain some value even if future SASIC (or even ASIC) devices are more efficient for mining.

OTOH, the density of one S6-LX150 per board is kinda poor, as I assume these are losing a fair amount to commons and PCB costs— (the bare FPGAs are about $190 each IIRC).  Sort of a tossup, the single chip boards are more reusable but it's still kinda pricy even with the quantity discounts.

[amazingrando]

You're right.    Sorry about that.  I pulled the wrong numbers so the output was too high for both FPGA and GPU.  I have updated the calculations, which now shows that FPGA and GPUs are roughly equivalent at 8 cents/kWh, everything else being equal.

People should run their own numbers in any case— your $/MH on the GPU farm is higher than my actuals by about 25% but people build with different degrees of efficiency.

You're certainly making the point that FPGAs are now in the realm of competitive for new deployments, even for buildouts with fairly short term planning horizons. Ztex's modules are also attractive because they've clearly been designed in a manner which would facilitate re-purposing them for things other than mining, which is a reason they should retain some value even if future SASIC (or even ASIC) devices are more efficient for mining.

OTOH, the density of one S6-LX150 per board is kinda poor, as I assume these are losing a fair amount to commons and PCB costs— (the bare FPGAs are about $190 each IIRC).  Sort of a tossup, the single chip boards are more reusable but it's still kinda pricy even with the quantity discounts.

25% less on your build?  That's impressive.  Care to share your secrets?

Very true:  Do you own math/homework. Whether it's hype, anecdotes, or my calculations, it might not be right for you.

I'm ambivalent about the resale value of the FPGAs. If BTC was no longer viable or you just wanted to get out of the game, I think you would take a much bigger hit on the residual value of the FPGAs than a GPU.

Certainly they are being an option worth serious consideration.  Multi-FPGA and better hash rate at a lower cost is what could be very interesting.  Keen to hear about the 2nd gen....

[gmaxwell]

25% less on your build?  That's impressive.  Care to share your secrets?

(sorry for the OT tangent)

5830s overclocked (and carefully clock managed, but not overvoltaged, 324MH/s) bought in bulk, 6x to a board, slowest cpu, network booting, a lot of careful tweaking. All other parts cut to a minimum. GPUs screwed to a wooden frame many linear feet long spaced as wide as possible. Ribbons ordered from hong-kong and shipped on the slow boat. I'm also fairly power efficient for a gpu setup— you can gain a fair bit by careful psu selection and running at 240v.  (and good power quality is important for stability at high clock rates)

Total cost per node are 109*6 (gpus) + 210 (mb) + 4*6 (ribbons) + 60 (cpu/ram)  + 100 (power) + 8 (marginal price of a ethernet switch port) = $1056 plus some modest shipping costs and a negligible amount of wood and screws.  yields 1944 MH/s.  Er. That a bit better than I said, though its a bit closer once you throw in some for shipping.

I'm ambivalent about the resale value of the FPGAs. If BTC was no longer viable or you just wanted to get out of the game, I think you would take a much bigger hit on the residual value of the FPGAs than a GPU.

Its hard to say. There are a lot of gpu miners... and once the cards are a couple generations old I expect an influx of used cards (especially a lot of lemon used cards that have been driven hard and not well cooled as many miners are guilty of) will undermine the used market price. It's hard to guess by how much.  These ATI GPUs are only so vastly superior to the alternatives for a few apps.. a lot of gamers would rather have nvidia.

If you want to build a DES cracker however, then that FPGA will be a pretty smoking solution.  I expect the FPGA vendors could probably improve their sales by offering tools for other applications, just to increase customer confidence that the FPGAs are useful for things other than mining.  DES reversing and WPA cracking are obvious things that come to mind. ... lack of fast IO limits SDR applications, lack of memory and fast IO limits imaging applications, these are best for crypto.