A custom designed FPGA miner for LTC?

[Nova!]

I have found what I believe is a shortcut in scrypt that if implemented correctly in hardware could dramatically speed up the hashrate.
I believe it should work and I know how I would implement it if I had the resources to acquire the FPGA and tools I need.

To show good faith I will elaborate on the algo and how the shortcut would work.  
This is really over simplified, but you are free to take this idea and roll with it.

scrypt the algo used by LTC and in fact all hashing algos, are comprised of 2 predominant steps.
#1 Generate a random list
#2 Hash across it.

To generate consistent results the random algo is actually deterministic pseudo-random and the setup for it is determined by a seed.
We will call this the prng.

The other step is hashing which is pretty well understood, you take a value from list a and replace it with a value from list b.
When you are done iterating you now have a hash.

scrypt differs mostly because it uses an entirely new list so frequently.  
The setup and tear down of this list requires quite a bit of CPU time and a lot of time is wasted on the memory bus performing storage & retrieval operations.
It cannot be done concurrently because the list itself changes frequently.

The shortcut is to have a multicore setup and a ton of on-die ram.
A dedicated prng core which does the setup and teardown for the second core.

The secondary core is the hashing core.  It would tell the prng core to setup a new list.
Then it would retrieve position x off the list from the shared memory space.
Other than that it would also perform all the normal hashing functions in a dedicated memory space.

I believe the total I need to make this work is about $12k USD, the FPGA I'm targeting right now is $10k and a license for the dev tools will be about $2k.
If I can find a less expensive option then I will go for that, but there aren't that many FPGAs that meet requirements right now.  
The particular target FPGA also has a direct path to ASIC from the mfr.

If you're willing to donate to the effort, I will keep you in the loop with full disclosure including build instructions and a copy of the sources and the firmware.
I haven't decided on a license for this if it works, but you will at least have a right to personal use.  
Perhaps if enough people are interested in production level manufacturing we could go a different route.  I'm not particularly interested in making this something I do for the rest of my life, but the contrarian in me is very excited by the potential here.

The LTC donation address is below.
LKfKkRMvMf2stQMNzQdKCvaf2YueAv1QSa

You can also donate BTC to the key in my sig.
There is no maximum but if you do decide to donate please send at least 0.5 LTC or the equivalent in BTC.
Then post just the address you donated from and I'll PM you here with a bitmessage key to join the group.

Thanks in advance!

[ryepdx]

Very few people will be willing to send you money on a promise. If you really want to do this, go the BkkCoins route and open source your efforts. Once people see there's a viable design that you've put effort into, they may be more willing to contribute funds. Otherwise you're probably going to have to bankroll the thing yourself and demonstrate a working product first.

[Nova!]

Very few people will be willing to send you money on a promise. If you really want to do this, go the BkkCoins route and open source your efforts. Once people see there's a viable design that you've put effort into, they may be more willing to contribute funds. Otherwise you're probably going to have to bankroll the thing yourself and demonstrate a working product first.

Yeah i understand that.  My other plan is to self bankroll it.  More time to save up but i keep the end result.
I would be doing this on my own anyways but i figured i would share and frankly 0.5LTC for everything you need for a miner except the chip is quite a bargain in my book.
I'll keep this thread up until my funding target is met either from contributions, my own efforts or likely some blend of the two.

If i find a cheaper fpga that will work and I haven't received any contributions by then, I'll just wrap the thread up and keep it close to my chest.
 

[fasmax]

A $10,000 dollar FPG chip wow. What do you think the hash rate would be?
Do you think it would beat my GPU?

[Magnate]

No r3eflection on yourself, but there are just soooo manyt scams around that nobody will trust anybody else without a show of proof.

I suggest getting a dev board, make you algo work even if highly restricted, then show the community how it will scale if you had more resources.

Dev board can be obtained for just a few hundred $$, clearly if you need special hardware setups you will need to get some funding through your own closer networks (your own $$, friends, family, put a call out to meet with LTC miners in your local area).

[Nova!]

A $10,000 dollar FPG chip wow. What do you think the hash rate would be?
Do you think it would beat my GPU?

Yes I'm pretty certain it would be more hashing power than most mining pools.

[Nova!]

No r3eflection on yourself, but there are just soooo manyt scams around that nobody will trust anybody else without a show of proof.

I suggest getting a dev board, make you algo work even if highly restricted, then show the community how it will scale if you had more resources.

Dev board can be obtained for just a few hundred $$, clearly if you need special hardware setups you will need to get some funding through your own closer networks (your own $$, friends, family, put a call out to meet with LTC miners in your local area).

Obviously I'm not running a scam.
I'm making an offer to to have a chance to help fund me so I can try this out.
The minimum works out to less than a dollar at current rates.  Most people spend more on satoshi dice.
In exchange you get to know everything I know, learn and discover as I go along.  
When the process is done, you get everything you need to build one yourself.

It's a $10,000 FPGA, that price includes the devboard. I know you guys are used to Xylinx and other lower end chips, but this is a chip designed for a completely different application it just has certain characteristics that make it an ideal miner.

However it is $10k.  Because it costs so dang much, I'm not going to try and build a community around it.  The only people who would be even interested at that price point are companies selling mining hardware and I really don't want to see this commercialized.  As the price comes down on the FPGA it's possible a community would crop up over time, but I really don't want to get married to a hardware project.
But it is a fascinating project and if nothing else at least folks are aware it's possible now at least in theory.

[efx]

So you are assuming no one else is working on this with a potentially large head start? Let's just say that's a little overly optimistic.

"Obviously I'm not running a scam. "

Mmmm, okay... I've heard that before. Even though you may not plan on scamming people, do you have the capitol to cover your investors?

Never heard of 'Xylinx'....perhaps the I is interchangeable.

Also, correct me if I'm wrong, but 'When the process is done, you get everything you need to build one yourself.' = paying another ~$10k if you aren't mass producing your own units, no?

[gica_contra]

So you are assuming no one else is working on this with a potentially large head start? Let's just say that's a little overly optimistic.

"Obviously I'm not running a scam. "

Mmmm, okay... I've heard that before. Even though you may not plan on scamming people, do you have the capitol to cover your investors?

Never heard of 'Xylinx'....perhaps the I is interchangeable.

Also, correct me if I'm wrong, but 'When the process is done, you get everything you need to build one yourself.' = paying another ~$10k if you aren't mass producing your own units, no?

I bet you've never heard of Adibas or Puna either. At 10k that is a very special chip...

[efx]

Um, have you heard of xylinx?  Recall that he implied xyilinx is a budget manufacturer.


 At 10k USD, it's damn sure not based on Xilinx arcs unless it's a huge cluster. Of course, that would be rubbish for sCrypt.

However, I would like to know what this magical $10k FPGA arc with large amounts of high bandwidth, low latency cache is actually called.

[gica_contra]

I was jesting. Those 2 are knock-off brands. Maybe the chip comes with a gold heat-sink. To match an entire pool it has to do some mighty hashing. Or maybe they've found a way to move electrons faster than the speed of light, going back in time and mining at a lower difficulty. After all, the chip sounds quite magic.

[Viceroy]

I m VERY interested.  tell me more.  I might fund it.

[efx]

Oh oops, my mistake, I gotcha now.  

The proposal is a touch too theoretical I'm afraid.

On balance, at least he isn't making wild efficiency claims that even a semi-competent electrical engineer would question (ahem BFL). lol


^ "tell me more" agreed.

[Nova!]

What investors?  I don't recall asking for investors at all.  I'm offering to share what I learn as I do this, period.  
If the information is of no value to you, then don't stress you're probably not the target audience.  
If it's something that piques your interest, pitch in.

Yes actually I'm fairly certain no one else is working on this and no one has a significant head start.  
I'm confident, because I've read what people are doing.  This is different, if you don't understand from my description how it's different then please re-read mine vs theirs.
I've also talked to the people who are working on FPGAs for litecoin they say it's intriguing but obviously they're targeting different hardware so this clearly wouldn't work for them.  

Point of fact at today prices, dollar for dollar you will get more hashing power with theirs.  That will probably be the case for a year or two.

This isn't about that, it's about a shortcut in scrypt that can be implemented in next gen FPGAs and which eliminates the memory bus problem.  
I'm saying I want to explore that particular shortcut, I don't have the resources to do the exploration, so if you want to chip in I'll let you know what I find.
Otherwise I will do it whenever I can.

Typo in the name of a mfr, sorry for that.  I'm tired it's 1:30 AM.  As long you understood who i meant, but just to clear up confusion http://www.xilinx.com/ good company, but doesn't make what I'm working on here.

You're not wrong, but you're assuming that the FPGA & devboard will stay the same price forever.  
My experience has been that these things go down and do so rather rapidly.  
Also it's not just the FPGA, a significant portion of that cost is in the board.
There is a bunch of stuff on the board that this project doesn't need, but it's on there anyways because it's a devboard.

When completed you would have all of the knowledge that I have attained in this endeavor.  
I'm only asking for donations because I'd like to move on this more quickly than my personal budget will allow.  
I am currently in a downtime situation where I have a couple months to throw full-time into this.

[Nova!]

I m VERY interested.  tell me more.  I might fund it.

I'll send you a PM in the morning, I need to get some sleep now.  The short of it is already up at the top and covers 3/4ths of the idea.  The other 1/4th is implementation specific details.

[efx]

Fair enough, but I would definitely consider it an investment. IP can be just as valuable as monetary returns if not more.  


I do understand your concept, actually one of the LTC devs already proposed the idea for 'setup and teardown' by a separate processor.

You are proposing both a shared and dedicated cache, correct? In essence, it would be similar to the gpu lookup gap function we see now, just without the dedicated co-processor. I would be interested in hearing the specifications of both caches and the link width (if you are attempting to avoid on-die cache like I think).

[Nova!]

I was jesting. Those 2 are knock-off brands. Maybe the chip comes with a gold heat-sink. To match an entire pool it has to do some mighty hashing. Or maybe they've found a way to move electrons faster than the speed of light, going back in time and mining at a lower difficulty. After all, the chip sounds quite magic.

It's an FPGA with a crap ton of on-die ram.  There is nothing magical about the chip.  By the time it comes down in price there will be better chips.
If I can find a less expensive one with that much ram tacked on, I will go with that instead.

On that point I'm open to suggestions.

[efx]

Oh I see I was mistaken. That's going to require both core density and a very large amount of the particularly expensive on-die cache, no? What process node is this based on? 


Are you looking to battle against 7990s and the like in both pure hashrate and efficiency?

[Nova!]

Fair enough, but I would definitely consider it an investment. IP can be just as valuable as monetary returns if not more.  


I do understand your concept, actually one of the LTC devs already proposed the idea for 'setup and teardown' by a separate processor.

You are proposing both a shared and dedicated cache, correct? In essence, it would be similar to the gpu lookup gap function we see now, just without the dedicated co-processor. I would be interested in hearing the specifications of both caches and the link width (if you are attempting to avoid on-die cache like I think).

Thanks I'll send you the same thing I do viceroy in the morning.  Need sleep now.

[efx]

Okay, thanks!