BOINC / F@H + FPGAs

[Cranky4u]

Is anyone thinking about developing a bitstream for CM1s to be retooled for use with BOINC or F@H?

Just thinkning ahead for ASIC release that I would like to retool my FPGAs to further scientific knowledge rather than collecting dust....My elec costs $0.25kW/hr so when BTC becomes uneconomical, insert ASIC release, then I would like to help humanity through science

[tytus]

F@H requires too many floating point operations and probably way too much RAM for FPGAs (very simple lattice folding models could be used but this is pure theory , no applicability). Smith Waterman (alignment, but preferably of nucleotide sequences not proteins) could run efficiently on FPGA. I see a project probably related to that on BOINC "Superlink@Technion" but I guess it would be much better to come up with our own project that fits FPGA much better. We could think about something like this and maybe even get some funding [probably realistic in EU or USA, but not cross continent].

Maybe we could discuss potential projects here.

[tytus]

I think a Gibbs sampler (http://bayesweb.wadsworth.org/gibbs/gibbs.html) would fit quite well on FPGA especially if it would deal with DNA sequences. We could then start looking for motifs in the nr database or in the human genome.

[Lethos]

Your biggest hurdle is that most alternative uses that potentially any FPGA can be re-purposed for, often requires a decent amount of fast on board memory to be of much use. None of the ones built for bitcoin needed this fast on-board memory, so were not built with any significant amount of them. So you have to consider that when you look into what it's going to be reused for.

I could be wrong, but projects suggested so far, are kind of memory hogs, how much that hurts it's ability to do it's job I don't know.

My FPGA's will be held onto till these are unaffordable to operate, which should take a while. Even then I will just store them for the day when they might be needed for bitcoin mining again. After all if the hashing algorithm ever changes, these FPGA's can be changed, ASIC's can not. I don't expect it to change any time soon if ever all, after all it would be pretty significant event and cripple all ASIC's in the process, more of a just incase scenario. It will take GPU's a while to get to the same level of energy-efficiency and we all know how quickly bitcoin has advance in 3 years.

[markm]

Bear in mind too that a huge percentage of the mining power currently deployed is totally, and in some cases quite deliberately, not merged-mining a full panoply of alternate blockchains.

Depending how many FPGAs are actually out there that would get on board with Massively Merged Mining, who knows how profitable some of the chains could become, and how much additional revenue all the chains that have had to go into hiding due to lacking enough hashing power to secure them could end up adding?

-MarkM-

[purelithium]

Bear in mind too that a huge percentage of the mining power currently deployed is totally, and in some cases quite deliberately, not merged-mining a full panoply of alternate blockchains.

Depending how many FPGAs are actually out there that would get on board with Massively Merged Mining, who knows how profitable some of the chains could become, and how much additional revenue all the chains that have had to go into hiding due to lacking enough hashing power to secure them could end up adding?

-MarkM-

That's a good point. We should all start pushing for folks to use Merged Mining Pools, ones that support a lot of the alternative coins. The one I'm on right now is great for it. (bitparking pool)

[Sitarow]

F@H requires too many floating point operations and probably way too much RAM for FPGAs (very simple lattice folding models could be used but this is pure theory , no applicability). Smith Waterman (alignment, but preferably of nucleotide sequences not proteins) could run efficiently on FPGA. I see a project probably related to that on BOINC "Superlink@Technion" but I guess it would be much better to come up with our own project that fits FPGA much better. We could think about something like this and maybe even get some funding [probably realistic in EU or USA, but not cross continent].

Maybe we could discuss potential projects here.

That would be something worth working towards.

[tytus]

Your biggest hurdle is that most alternative uses that potentially any FPGA can be re-purposed for, often requires a decent amount of fast on board memory to be of much use. None of the ones built for bitcoin needed this fast on-board memory, so were not built with any significant amount of them. So you have to consider that when you look into what it's going to be reused for.

1. Gibbs sampler (or other multiple sequence alignment generating methods) don;t need much memory
2. FPGA has block RAM and distributed RAM that would be sufficient for many applications

[Cranky4u]

Bear in mind too that a huge percentage of the mining power currently deployed is totally, and in some cases quite deliberately, not merged-mining a full panoply of alternate blockchains.

Depending how many FPGAs are actually out there that would get on board with Massively Merged Mining, who knows how profitable some of the chains could become, and how much additional revenue all the chains that have had to go into hiding due to lacking enough hashing power to secure them could end up adding?

-MarkM-

That's a good point. We should all start pushing for folks to use Merged Mining Pools, ones that support a lot of the alternative coins. The one I'm on right now is great for it. (bitparking pool)

I am keen to mine alternative cryptocurrencies however it is my understanding that to do this would require new bitstreams. Is this true or can the current BTC bitstreams be reused for other, currently exisiting, alt-currencies? IF so, which ones?

Cheers

[markm]

I am keen to mine alternative cryptocurrencies however it is my understanding that to do this would require new bitstreams. Is this true or can the current BTC bitstreams be reused for other, currently exisiting, alt-currencies? IF so, which ones?

Cheers

Do your FPGAs work with p2pool ? If so they should have no  trouble doing Massively Merged Mining.

-MarkM-

[Gabi]

To use the FPGA for boinc/f@h you need the proper bitstream. Problem, making the bitstream for bitcoin mining is fairly easy, it's just sha256 hashing. While making it for these distributed computing project is much much harder, because what they do is more complex.
Also projects update their client with new functions and so on, so you would have to update the bitstream when this happens.

[tytus]

You can use FPGA to solve partial problems of a more general problem. For example smith waterman can be applied in many scenarios.