A RAM based fpga LTC miner

[kmtan]

mark.

interesting to have 1..pls

[Hippie Tech]

Bzzzzzzzzzzz

[hendo420]

[BFL - CEO]

I'm not watching.

[DeathAndTaxes]

I'm not watching.

I am surprised BFL isn't already accepting pre-orders for a 1 MH/s "Litecoin Single" and 100 MH/s "Litecoin minirig" with delivery in 4-6 weeks (TM).

[smolen]

A million Quark question: where all the GPU hashpower will go when this thing will be released?

[DeathAndTaxes]

A million Quark question: where all the GPU hashpower will go when this thing will be released?

We don't know if this thing will be released.  It may not be economical yet.  It is an inevitability that dedicated devices to mine LTC (and clones) will eventually be created though.   The creators made sure of that by crippling it's memory hardness (LTC 128KB max scratchpad vs Scrypt default 16MB)  I would imagine when that happens most GPUs will end up on ebay.

[smolen]

It may not be economical yet.

Seems this design employs several independent memory buses, so the price of FPGAs will be measured not in LUT/$, but in pin/$. Great time for people with access to used components market.

[co5hike]

Seems like fun project.

I waiting for numbers, but I guess GPUs are safe for now

[Hydroponica]

…Looks at LTC difficulty, without existence of Scrypt Asics....

…Looks at current LTC value, without existence of Scrypt Asics....

…Idiot

[gdassori]

Yet another FPGA with RAM.
What's new here ?

I have seen tons of pics like this in the last 3 months, and today I still can't say STFU AND TAKE MY MONEY.

[BFL - CEO]

Yet another FPGA with RAM.
What's new here ?

I have seen tons of pics like this in the last 3 months, and today I still can't say STFU AND TAKE MY MONEY.

You've seen tons of pictures like this?  Can you point me to those pictures because I've never seen a picture like the above.

[Bonz]

…Looks at LTC difficulty, without existence of Scrypt Asics....

…Looks at current LTC value, without existence of Scrypt Asics....

…Idiot

this would be an FPGA not a ASIC!  FPGA's didn't kill the bitcoin world and they won't kill the litecoin world either

[gdassori]

BFL CEO: U have PM, I'm not here to advertise something.

By the way, I haven't said nothing against this, I'm just saying "where is the news? It has achieved good results? There are some noteworthy hashrate?", a PIC of an FPGA with DRAM means nothing.

[Pt0x]

I hope to see a good hash rate coming out  from this device!

[marnem]

watching

[yochdog]

The next great unicorn....

[SaltySpitoon]

I'm aware that this is a FPGA which is doable with Scrypt, however I'd like to go off in a minor tangent. People seem to underestimate how difficult it will be to create a Scrypt ASIC. SHA256 Asics have been used for many many years. They were not new technology, meaning the billions of dollars of research that others had done getting SHA256 ASICs working is not there already for proposed Scrypt ASICs. All the BTC mining ASIC companies needed to do, was make a product that would work for BTC specific hashing, rather than what they were and still are used for, encrypting and decrypting files. The company that decides to start making LTC Asics will need a whole lot more than a few hundred thousand BTC to get their products out the door.

Back on topic, LTC FPGAs actually aren't that difficult to make in theory. LTC's Scrypt hashing requires actually a much lower amount of memory than other scrypt implementations (I believe its 196mb/cycle although I may be off) at that point, or whatever it actually is, I remember the math behind it, but not the actual numbers, you can provide additional hashing power at 1/2 the memory required, and you can still end up with a higher hashrate over current GPUs, while still using fairly inexpensive FPGA technology. So rather than needing to create a new FPGA board that can handle uneconomical amounts of memory, you can just work on designing a chip that will hash fast, and lose performance based on how much memory you can actually supply.

I'll look back over my research tomorrow, and get all of the numbers and such down. I'm tired so I may have said something dumb, I'll correct it later.

[DeathAndTaxes]

I'm aware that this is a FPGA which is doable with Scrypt, however I'd like to go off in a minor tangent. People seem to underestimate how difficult it will be to create a Scrypt ASIC. SHA256 Asics have been used for many many years. They were not new technology, meaning the billions of dollars of research that others had done getting SHA256 ASICs working is not there already for proposed Scrypt ASICs. All the BTC mining ASIC companies needed to do, was make a product that would work for BTC specific hashing, rather than what they were and still are used for, encrypting and decrypting files. The company that decides to start making LTC Asics will need a whole lot more than a few hundred thousand BTC to get their products out the door.

Back on topic, LTC FPGAs actually aren't that difficult to make in theory. LTC's Scrypt hashing requires actually a much lower amount of memory than other scrypt implementations (I believe its 196mb/cycle although I may be off) at that point, or whatever it actually is, I remember the math behind it, but not the actual numbers, you can provide additional hashing power at 1/2 the memory required, and you can still end up with a higher hashrate over current GPUs, while still using fairly inexpensive FPGA technology. So rather than needing to create a new FPGA board that can handle uneconomical amounts of memory, you can just work on designing a chip that will hash fast, and lose performance based on how much memory you can actually supply.

I'll look back over my research tomorrow, and get all of the numbers and such down. I'm tired so I may have said something dumb, I'll correct it later.

LTC uses the parameters (2^10, 1, 1) which results in a token 128KB max scratchpad size.  That isn't a typo it is kilobytes.  The default Scrypt parameters (2^14, 8, 1) result in a 16MB max scratchpad size roughly 128x as "memory hard". 

To my knowledge no Bitcoin ASIC company used existing SHA-2 IP and modified it.   
 

[SaltySpitoon]

I'm aware that this is a FPGA which is doable with Scrypt, however I'd like to go off in a minor tangent. People seem to underestimate how difficult it will be to create a Scrypt ASIC. SHA256 Asics have been used for many many years. They were not new technology, meaning the billions of dollars of research that others had done getting SHA256 ASICs working is not there already for proposed Scrypt ASICs. All the BTC mining ASIC companies needed to do, was make a product that would work for BTC specific hashing, rather than what they were and still are used for, encrypting and decrypting files. The company that decides to start making LTC Asics will need a whole lot more than a few hundred thousand BTC to get their products out the door.

Back on topic, LTC FPGAs actually aren't that difficult to make in theory. LTC's Scrypt hashing requires actually a much lower amount of memory than other scrypt implementations (I believe its 196mb/cycle although I may be off) at that point, or whatever it actually is, I remember the math behind it, but not the actual numbers, you can provide additional hashing power at 1/2 the memory required, and you can still end up with a higher hashrate over current GPUs, while still using fairly inexpensive FPGA technology. So rather than needing to create a new FPGA board that can handle uneconomical amounts of memory, you can just work on designing a chip that will hash fast, and lose performance based on how much memory you can actually supply.

I'll look back over my research tomorrow, and get all of the numbers and such down. I'm tired so I may have said something dumb, I'll correct it later.

LTC uses the parameters (2^10, 1, 1) which results in a token 128KB max scratchpad size.  That isn't a typo it is kilobytes.  The default Scrypt parameters (2^14, 8, 1) result in a 16MB max scratchpad size roughly 128x as "memory hard". 

To my knowledge no Bitcoin ASIC company used existing SHA-2 IP and modified it.   
 

you are correct, I was thinking it was 196kb for some reason (as mentioned tired) all Bitcoin ASIC companies had to derive their works from existing the current SHA-2 ASICs, starting from scratch would have cost far more than the Bitcoin economy could have supplied, and far more than ASIC companies could afford to spend at their current price points. Its like if current ASIC companies decided to start using 14nm chips. It would be unimaginably expensive to create a technology that doesn't exist yet, its far cheaper to modify existing designs. I haven't actually sat down and talked to the ASIC manufacturers, but I'd say its a pretty strong gut feeling.

I've got some super secret projects that would be neat if I could run by you tomorrow (ok not that super secret). I'm at the point where I have to read over my posts 30 times to make sure I spelled everything rigt, and should probably get some sleep.