Mining with an AVNET Spartan 6 FPGA LX9 Microboard

[lucasjkr]

A friend of mine just provided me with an LX9 FPGA Board. No cost to myself. Is this thing worth mining with? I read in one place that a Spartan 6 could be expected to generate 100 MH/s, but read somewhere else than that's for a beefier board, and that an LX9 might only generate 3 or 5 MH per second?

Anyone have a clue about this?

[OnkelPaul]

Search the forum for LX9.
Some old discussion: https://bitcointalk.org/index.php?topic=9047.msg472810#msg472810

This is certainly a nice little fun device, but hashing power is too limited.

Onkel Paul

[BitcoinFX]

These are mostly useless for Bitcoin mining now.

However, I'm actually working on a project for a USB FPGA miner for scrypt based coins.

I'm using one of these Xilinx® Spartan®-6 FPGA LX9 MicroBoards: http://www.xilinx.com/products/boards-and-kits/AES-S6MB-LX9.htm

with 2 added PmodSF2 - Serial PCM, 128 Mbit: http://digilentinc.com/Products/Detail.cfm?NavPath=2,401,840&Prod=PMOD-SF2

This is a personal project atm and I'm having to design / build software to get these working fully to try and overcome the difficulties of scrypt mining being very memory intensive. If / when I can get these hashing scrypt properly then I may offer them for sale with the software as custom built scrypt mining kits. Running 4 sticks with an Rpi might present an opportunity for ROI vs electricity usage and purchase costs etc.

However, to get these hashing scrypt at a reasonable and stable throughput, the way that I've currently found to improve performance requires around 4GB+ of free RAM on the host machine for one FPGA (and around 2GB+ for each additional FPGA). I'll have more info. / updates after doing much more testing in my free time over the coming weeks / months. This might well just end up being another fun project though.

[lucasjkr]

Yes, that's the unit i have.

Any idea of what the hashing rate will be under scrypt? Trying to figure out if i should just give it back or if there's any point to me (a non-programmer) having one of these?

[J35st3r]

Any idea of what the hashing rate will be under scrypt? Trying to figure out if i should just give it back or if there's any point to me (a non-programmer) having one of these?

Its going to be pathetic, even if he gets it to work at all! You need memory (lots of it) for scrypt, and there is no way of attaching any to that little usb board, so it will rely on the host over the usb connection. Its going to be slower than CPU mining.

It will do bitcoin mining though (at about 5MHash/sec using the standard fpgaminer code), so it will earn you a couple of USD cents a day, less the electricity cost for running your host computer, so that's a money sink by my reckoning, even if you hosted it off something low-powered like a raspberry pi (which will need some custom software coded for the usb interface as the xilinx toolset certainly won't run on it).

My advice, send it back and get a proper miner (not one of those ASIC pre-order scams though, you'll never get your money back by the time you eventually take delivery).

Just my 2c BTC

[BitcoinFX]

Any idea of what the hashing rate will be under scrypt? Trying to figure out if i should just give it back or if there's any point to me (a non-programmer) having one of these?

Its going to be pathetic, even if he gets it to work at all! You need memory (lots of it) for scrypt, and there is no way of attaching any to that little usb board, so it will rely on the host over the usb connection. Its going to be slower than CPU mining.

It will do bitcoin mining though (at about 5MHash/sec using the standard fpgaminer code), so it will earn you a couple of USD cents a day, less the electricity cost for running your host computer, so that's a money sink by my reckoning, even if you hosted it off something low-powered like a raspberry pi (which will need some custom software coded for the usb interface as the xilinx toolset certainly won't run on it).

My advice, send it back and get a proper miner (not one of those ASIC pre-order scams though, you'll never get your money back by the time you eventually take delivery).

Just my 2c BTC

Yep. All 100% correct and the speeds are going to be pretty lame indeed. You will get much better hashing speeds with good multi core CPU's and even a low-to-mid range (much cheaper) ATI 5xxx card for the price.

I'm mostly doing this for the fun of it, as I already have this board and it makes of an interesting add-on project for a PC that's hashing to the max. with it's CPU's and GPU's already. I'm looking to do something with all that free system RAM.

Knowing what might be possible and having experience with these FPGA's and developing code for scrypt mining will be applicable to better / faster FPGA boards in the future. Hopefully that explains why I'm even bothering with this.

[xh4]

with 2 added PmodSF2 - Serial PCM, 128 Mbit: http://digilentinc.com/Products/Detail.cfm?NavPath=2,401,840&Prod=PMOD-SF2

I think you may run into some bottleneck issues because of SPI, but if the hashrate is low enough it might not matter...

[BitcoinFX]

with 2 added PmodSF2 - Serial PCM, 128 Mbit: http://digilentinc.com/Products/Detail.cfm?NavPath=2,401,840&Prod=PMOD-SF2

I think you may run into some bottleneck issues because of SPI, but if the hashrate is low enough it might not matter...

Yes for sure. Could probably try limiting the hash rate to make it more efficient by avoiding overflow, if that makes sense.

Someone made a useful post on the Litecoin forum - FPGA Development thread

See: https://forum.litecoin.net/index.php/topic,2702.msg22071.html#msg22071

Anyway, I got this board hashing Bitcoin via a Linux box, started doing some Litecoin / scrypt coding and testing and messed up my installation.

Links for historical projects I've found useful include:

Open Source FPGA Bitcoin Miner : https://github.com/progranism/Open-Source-FPGA-Bitcoin-Miner

Xilinx Serial Miner : https://github.com/teknohog/Xilinx-Serial-Miner

Fork : https://github.com/teknohog/Open-Source-FPGA-Bitcoin-Miner

cgminer and bfgminer ofc and the original scrypt paper is obviously worth a look: https://www.tarsnap.com/scrypt/scrypt.pdf

I'd be interested to here from anyone who has already got this LX9 microboard hashing Bitcoin in Windows.

The Scrypt FPGA concept is becoming more popular: https://forum.litecoin.net/index.php/topic,2725.0.html

I'll stick with my homebrew project for now.  

[J35st3r]

Ah, sorry, I missed the bit about the SPI memory. I also see it has 64MB of sdram on board, you won't be relying on the host's ram after all.

I did consider having a punt at scrypt on my DE0-Nano which also has some sdram on board, but I reckoned the throughput would only be about 60 MByte/sec in burst mode (or 200MBytes/sec if I'm mis-calculating the precharge latency ... I don't have the docs to hand right now and I'm not going to go looking it up just for this little post). Anyway you've got a 100kB lookup table to build and swizzle for each hash, so I reckoned the resulting hash rate wouldn't be worth the huge number of J35st3r-hours to get it working). A fun looking project though, and an A* for your effort. Good luck.

[WindMaster]

This is a personal project atm and I'm having to design / build software to get these working fully to try and overcome the difficulties of scrypt mining being very memory intensive. If / when I can get these hashing scrypt properly then I may offer them for sale with the software as custom built scrypt mining kits. Running 4 sticks with an Rpi might present an opportunity for ROI vs electricity usage and purchase costs etc.

ROI for scrypt mining is extremely unlikely with an XC6SLX9 FPGA.

However, to get these hashing scrypt at a reasonable and stable throughput, the way that I've currently found to improve performance requires around 4GB+ of free RAM on the host machine for one FPGA (and around 2GB+ for each additional FPGA). I'll have more info. / updates after doing much more testing in my free time over the coming weeks / months. This might well just end up being another fun project though.

If your goal is to satisfy the scrypt memory requirements by placing the buffers in system RAM and sending data back and forth over USB, this is very definitely the wrong approach.  And not just by a little bit!  Just run the calculation for USB bulk endpoint throughput vs. how many bytes you need to read/write to calculate a hash and you'll see why.

[teknohog]

Links for historical projects I've found useful include:

Open Source FPGA Bitcoin Miner : https://github.com/progranism/Open-Source-FPGA-Bitcoin-Miner

Xilinx Serial Miner : https://github.com/teknohog/Xilinx-Serial-Miner

Fork : https://github.com/teknohog/Open-Source-FPGA-Bitcoin-Miner

cgminer and bfgminer ofc and the original scrypt paper is obviously worth a look: https://www.tarsnap.com/scrypt/scrypt.pdf

While you're referring to all these open source projects, care to share some of your scrypt mining code? I'm not very active in this scene at the moment, but it would be interesting to try this on different FPGAs...

[BitcoinFX]

Links for historical projects I've found useful include:

Open Source FPGA Bitcoin Miner : https://github.com/progranism/Open-Source-FPGA-Bitcoin-Miner

Xilinx Serial Miner : https://github.com/teknohog/Xilinx-Serial-Miner

Fork : https://github.com/teknohog/Open-Source-FPGA-Bitcoin-Miner

cgminer and bfgminer ofc and the original scrypt paper is obviously worth a look: https://www.tarsnap.com/scrypt/scrypt.pdf

While you're referring to all these open source projects, care to share some of your scrypt mining code? I'm not very active in this scene at the moment, but it would be interesting to try this on different FPGAs...

Of course teknohog (thanks for sharing your code btw) my progress is slow at present, although when I've got something working well it will certainly go up on github. Share and share alike and very much still tinkering with this.