Bitcoin Forum

This is an open source quad spartan-6 LX150 board designed by The Seven and myself in collaboration.

Since it was completed way behind schedule we didn't see much point in releasing this version.
However some people kept asking for open source boards recently so we figured why not.

Version 1.0 has been manufactured and tested, but rev 2.1 has not. The differences are however minor.
The software is not very complete nor user friendly at the moment but The Seven has been mining with this board for a while now.
The plan is to make a  28nm version of the P7 platform where the software will include support for the LX150 version as well.

We will try to give as much support as possible through this forum and IRC, but the files are delivered as
is with no warranty of any kind. Use at your own risk.

Read the txt file included for more information.

~Pusle


download link:https://mega.co.nz/#!PcNlnBKJ!PN44_C7M2L0aiEd8BxyZNEbpGtJoytkXHbfSBkTAdFQ (https://mega.co.nz/#!PcNlnBKJ!PN44_C7M2L0aiEd8BxyZNEbpGtJoytkXHbfSBkTAdFQ)

http://oi44.tinypic.com/2vvl3qr.jpg

nice work.

So what happened to the diff clock driver, did you ever get it to work as expected?

Download link doesn't work for me. MEGA demands I should downgrade to Firefox browser while I'm using SeaMonkey.

Same link won't work for me either :(

Use Google Chrome. File size is 81MB

Yes, it's not a perfect square but it works nicely.  
I've tested many different drive strengths and termination resistors which didn't have much effect on overclocking.
There seems to be a wide margin.

The included files are set to LVDS 2.5V for the synth, which is the recommended setting
For extreme overclockers you may get a little more speed using SSTL 3.3 as it's even more square and faster rise/fall time.
However with SSTL the signal swing is so large it's way out of spec and not recommended.

Thanks, got it.

Can you share some more information with us? Hashing speed, watt usage, estimated price, drivers etc? It would be nice, if it will turn into a "product" which is able to sell. But really nice work anyway.

I guess the link should be fixed by enclosing it in [ url ] ... [ /url ] tags, the forum software doesn't seem to deal with anchors correctly.

Check the production files, and found some more photos of the project here we go! Seems more than impressive - the files, not the pictures :)  https://plus.google.com/photos/109390909883687914366/albums/5882375517910208881?authkey=CNmwoJrUiKWd6AE

Would be very interested to know hashrate/price/power of the 28nm version. I suspect it would work a lot better than current generation ASICs.

I assume Hashing speed and watt usage will be very similar to ztex 1.15y since its the same IC. Very small variations between different implementations of LX150

Someone do group manufacture/assembly please.  ;)

I'd mine the heck out of Terracoins/Ppcoins if I could get some of these cheap.

Maybe some Bitcoins along the way... but that's what I'm buying ASICs for..  :)

Very cool stuff, pusle and TheSeven!

It won't get close to current generation ASICs, neither in power nor cost.

With the current open source designs for Kintex-7, a quad board would get ~2GH/s, though I suspect 4GH/s is possible when more hashing cores are rolled into the design.  Power would be less than 100W at 2GH/s.  Price?  Well K160T chips are ~90$USD each so maybe a BOM of 800$USD?

ASICMINER blade is for $6200 (49.9 BTC). @10 GH/s
If BOM of $800 so total cost $1000 is sane assumption? Including assembly, pcb, etc?
Would need 2.5 to 5 (4 x K160T) boards for 10 GH/s ~$2500 - $5000

Power is another story... ASIC way way more efficient.

If someone can build this, and the $/GH is same as ASICMINER blades(or whatever ASIC unit is shipping at the time) im sure a lot of people would buy irrespective of power usage simply because the barrier to entry is lower....

but maybe by the time someone manufactures these... Avalon DYI projects churn out much cheaper $/GH..

Nice piece of hardware :)


+1

nice work.

That's correct. The current fpgaminer implementation is only at 180MHz but 200 should be reachable with some Xilinx tool massaging.
Most LX150 will overclock above 200MHz on this board giving a total hashrate above 800MH/s even with the included bitstream.

~Pusle

The software controllable VCCint voltage for the FPGAs might allow this to be increased even further, as you can easily just run the chips at 1.28V :)
(Or at lower voltages to increase power efficiency.)

When will you have Kintex-7 XC7K325T-1FB676C based design ready?
I can get them very cheap $150 to $110 based on quantity.

From the prices given by the local distributor it seems K160T-1FBG484 gives the most bang for the buck.
The K325T costs about two times more and it's easier to deal with a smaller package.

There is no schedule yet for the 28nm version.

how much megahash per second doing this board?

Half the price, half the logic -> probably half the hash speed -> what's the point?

"two times more" would translate into "a third of the price"...

hi

do you have the vhdl for this one?

So, if I follow you correctly, you would advise getting a device 50% slower to save 30% on price. I don't find that enticing.

To pay 30% of the price!

getting 25k slices and 600DSP blocks for 100$ is better than 50k slices and 840DSP blocks for 200$.
The package is cheaper too so the price might be even better in favor of the K160 and higher volume = lower price.

Another benefit with a smaller package is easier and simpler pcb layout and it hurts less for each part that doesn't work due to bad soldering etc

K160 ≃ 90USD/100pcs
K325 ≃ 110-150USD /?

Where do you get the 70% reduction in cost? (pcb + 4 xilinx -> ? -> -70%)

quote for the K325 in 100pcs was 187USD

right, but the maths still don't show a 70% reduction. you get a bit more than 50% reduction for the kintex but you still have the pcb cost, so you save less than 50%.

anyway...

very nice work :) I'm a big fan of the x6500 team and was hoping some of you would get back into making gear again. Thank you for releasing this as an open source project!