[ANN] Fibonacci Info and questions

[WindMaster]

Sorry if this has been asked before, but will it work with Scrypt-Jane, such as YaCoin?
Thanks,
Flippyfeet

Been a bit busy, I will ask my engineer to be sure. But looking at what jane is (Chacha) it probably would yes.

Negative.  The logic would be radically different between a Litecoin-oriented Salsa20/8 (1024,1,1) implementation and a YACoin-oriented ChaCha20/8 (N,1,1) + Keccak512 implementation.  There is very little similarity between the scrypt variant used in Litecoin and the scrypt variant used in YACoin.  Additionally, YACoin uses variable N while Litecoin has fixed N=1024.  For YACoin, N is currently 32768, and will change to 65536 on 31 May 2014.  Details on the variable N schedule can be found in my post here:

https://bitcointalk.org/index.php?topic=206577.msg2162620#msg2162620

When YACoin launched at N=32, I adapted 4 of our earlier FPGA LTC prototypes (populated with XC7A200T's and a pile of DDR3) to YACoin instead.  It was real easy pickings at N=32.  At N=64, not so much.  At N=128, I stopped and moved back to a different more efficient hardware solution for YACoin mining.

Variable N would make ASIC implementation a real bitch, since you can't optimize the mixing function's logic for a specific value of N like you can with Salsa20/8 (1024,1,1).

[1715118149]

1715118149

[1715118149]

1715118149

[1715118149]

1715118149

[WindMaster]

It is, and I know it would take someone months or even 1+ years to reverse engineer it. But I know what our teams values are, and I know (after dealing with lots of potential VCs) other companies have different values. So I feel more comfortable making them work hard to beat/meet our numbers rather than giving them the answers.

Just playing devil's advocate here, but I'd tend to say that a team capable of reverse-engineering an ASIC by selectively etching layers and taping-out a replica, already has the skillset to just develop an scrypt core from scratch.  Given that scrypt's internal workings aren't secret knowledge, and the possible hardware approaches to go about it are generally self-evident and well-documented from scrypt research outside of the cryptocurrency community, I think it would take less time to implement from scratch rather than reverse someone else's design.  If we were talking about replicating, say, Sandy Bridge, or some of AMD and Nvidia's GPU dies, that would be one thing.  But an ASIC implementation of scrypt just isn't that hard to do from scratch.  95% of other cores that end up in various ASIC's across other industries are significantly more complex than an scrypt core.

I'd say any team with the knowledge needed to reverse an ASIC, can probably also easily determine the appropriate TMTO and SRAM size (or external DDR3/GDDR5 size) for the particular silicon process node they're targetting to optimize cost/performance (and, really, there's only 2 choices along the TMTO spectrum that are going to land near the peak cost/performance for an ASIC implementation with on-die SRAM, and only 1 optimal TMTO choice for external DDR3 or GDDR5).  I don't see where there's much to gain in anyone's ASIC development cycle from looking at someone else's scrypt die.

[jasinlee]

It is, and I know it would take someone months or even 1+ years to reverse engineer it. But I know what our teams values are, and I know (after dealing with lots of potential VCs) other companies have different values. So I feel more comfortable making them work hard to beat/meet our numbers rather than giving them the answers.

Just playing devil's advocate here, but I'd tend to say that a team capable of reverse-engineering an ASIC by selectively etching layers and taping-out a replica, already has the skillset to just develop an scrypt core from scratch.  Given that scrypt's internal workings aren't secret knowledge, and the possible hardware approaches to go about it are generally self-evident and well-documented from scrypt research outside of the cryptocurrency community, I think it would take less time to implement from scratch rather than reverse someone else's design.  If we were talking about replicating, say, Sandy Bridge, or some of AMD and Nvidia's GPU dies, that would be one thing.  But an ASIC implementation of scrypt just isn't that hard to do from scratch.  95% of other cores that end up in various ASIC's across other industries are significantly more complex than an scrypt core.

I'd say any team with the knowledge needed to reverse an ASIC, can probably also easily determine the appropriate TMTO and SRAM size (or external DDR3/GDDR5 size) for the particular silicon process node they're targetting to optimize cost/performance (and, really, there's only 2 choices along the TMTO spectrum that are going to land near the peak cost/performance for an ASIC implementation with on-die SRAM, and only 1 optimal TMTO choice for external DDR3 or GDDR5).  I don't see where there's much to gain in anyone's ASIC development cycle from looking at someone else's scrypt die.

That is good then no one will be upset that we dont display it to the world 

We know what we have, there are a few community members that are aware too. All of them agreed this was the best approach, it was actually suggested by one of the forum admins in the first place to host and not ship immediately.

[jasinlee]

Sorry if this has been asked before, but will it work with Scrypt-Jane, such as YaCoin?
Thanks,
Flippyfeet

Been a bit busy, I will ask my engineer to be sure. But looking at what jane is (Chacha) it probably would yes.

Negative.  The logic would be radically different between a Litecoin-oriented Salsa20/8 (1024,1,1) implementation and a YACoin-oriented ChaCha20/8 (N,1,1) + Keccak512 implementation.  There is very little similarity between the scrypt variant used in Litecoin and the scrypt variant used in YACoin.  Additionally, YACoin uses variable N while Litecoin has fixed N=1024.  For YACoin, N is currently 32768, and will change to 65536 on 31 May 2014.  Details on the variable N schedule can be found in my post here:

https://bitcointalk.org/index.php?topic=206577.msg2162620#msg2162620

When YACoin launched at N=32, I adapted 4 of our earlier FPGA LTC prototypes (populated with XC7A200T's and a pile of DDR3) to YACoin instead.  It was real easy pickings at N=32.  At N=64, not so much.  At N=128, I stopped and moved back to a different more efficient hardware solution for YACoin mining.

Variable N would make ASIC implementation a real bitch, since you can't optimize the mixing function's logic for a specific value of N like you can with Salsa20/8 (1024,1,1).

Thank you for the clarification windmaster.

[Dabs]

What would you need to ship immediately instead of host? I'm assuming if you can trust who you send to, to not reverse engineer it. Like, for example, uh, escrow agents? hehehe.

*edit* Also, I won't sell it to anyone else until you start shipping it to the rest of the world.

And, does your discount only apply to LTC? Can we get the same discount for BTC? (For those who don't want to spend their LTC, or for those who don't have LTC, to prevent them from having to buy on an exchange.)

[eXme]

What would you need to ship immediately instead of host? I'm assuming if you can trust who you send to, to not reverse engineer it. Like, for example, uh, escrow agents? hehehe.

*edit* Also, I won't sell it to anyone else until you start shipping it to the rest of the world.

And, does your discount only apply to LTC? Can we get the same discount for BTC? (For those who don't want to spend their LTC, or for those who don't have LTC, to prevent them from having to buy on an exchange.)

lol, if he said LTC that means LTC not BTC or other coin. you smoke pot?

[Dabs]

What would you need to ship immediately instead of host? I'm assuming if you can trust who you send to, to not reverse engineer it. Like, for example, uh, escrow agents? hehehe.

*edit* Also, I won't sell it to anyone else until you start shipping it to the rest of the world.

And, does your discount only apply to LTC? Can we get the same discount for BTC? (For those who don't want to spend their LTC, or for those who don't have LTC, to prevent them from having to buy on an exchange.)

lol, if he said LTC that means LTC not BTC or other coin. you smoke pot?

Nope. Not in a long time. Several years ago. (My forum nick has nothing to do with pot, it's a coincidence.)

I know he said LTC, but maybe they might give a discount to those who will pay with BTC. If not, then I will simply buy LTC using my BTC and pay that way. The miners will still be pegged to it's fiat cost.

[eXme]

What would you need to ship immediately instead of host? I'm assuming if you can trust who you send to, to not reverse engineer it. Like, for example, uh, escrow agents? hehehe.

*edit* Also, I won't sell it to anyone else until you start shipping it to the rest of the world.

And, does your discount only apply to LTC? Can we get the same discount for BTC? (For those who don't want to spend their LTC, or for those who don't have LTC, to prevent them from having to buy on an exchange.)

lol, if he said LTC that means LTC not BTC or other coin. you smoke pot?

Nope. Not in a long time. Several years ago. (My forum nick has nothing to do with pot, it's a coincidence.)

I know he said LTC, but maybe they might give a discount to those who will pay with BTC. If not, then I will simply buy LTC using my BTC and pay that way. The miners will still be pegged to it's fiat cost.

you could be so smart 2 post ago then there would be any conversation about it 

[jasinlee]

Ok lets chill, it will only be in ltc.

[atp1916]

I woudn't even consider letting these things out into the wild until viable competing devices show up.

If you folks truly have what you say you have, i hope you have a serious fab contract. 

[Dabs]

I'll still buy 10 please. I'll just convert my different coins to LTC then.

[jasinlee]

I woudn't even consider letting these things out into the wild until viable competing devices show up.

If you folks truly have what you say you have, i hope you have a serious fab contract. 

We battled with ideas on how to handle this for about 4 months, the distributed hosting suggestion came from outside our team and I must say it was genius.

[atp1916]

I woudn't even consider letting these things out into the wild until viable competing devices show up.

If you folks truly have what you say you have, i hope you have a serious fab contract.  

We battled with ideas on how to handle this for about 4 months, the distributed hosting suggestion came from outside our team and I must say it was genius.

Let me ask you this: did it come from BinaryTriStar?

He was going to build an FPGA for us, and that was exactly his suggestion to keep the 1st gen units in-house until the 2nd/3rd gen stuff came out.

Either way, he completely dissapeared but that's old, dead history.  Just was curious.

[jasinlee]

I woudn't even consider letting these things out into the wild until viable competing devices show up.

If you folks truly have what you say you have, i hope you have a serious fab contract.  

We battled with ideas on how to handle this for about 4 months, the distributed hosting suggestion came from outside our team and I must say it was genius.

Let me ask you this: did it come from BinaryTriStar?

He was going to build an FPGA for us, and that was exactly his suggestion.

Either way, he completely dissapeared but that's old, dead history.  Just was curious.

You know, someone mentioned that name on the litecoin forum but I had never actually seen him post on here.

And no the suggestion came from a forum admin, I will ask him permission before saying it was his idea

[atp1916]

I woudn't even consider letting these things out into the wild until viable competing devices show up.

If you folks truly have what you say you have, i hope you have a serious fab contract.  

We battled with ideas on how to handle this for about 4 months, the distributed hosting suggestion came from outside our team and I must say it was genius.

Let me ask you this: did it come from BinaryTriStar?

He was going to build an FPGA for us, and that was exactly his suggestion.

Either way, he completely dissapeared but that's old, dead history.  Just was curious.

You know, someone mentioned that name on the litecoin forum but I had never actually seen him post on here.

And no the suggestion came from a forum admin, I will ask him permission before saying it was his idea

No big deal.  I was just curious.

BinaryTriStar was gonna dev a blade design here for us but he completely dissapeared after saying he had a prototype up and running!

[jasinlee]

Thats messed up, I would think there were more of those kinda schemes out there. I have not really seen anyone straight up pull a scam yet for scrypt miners. I guess he was the first. There are a bunch of companies popping up, and their speeds and specs seem to be very close to the datasheet we posted. I think this kinda thing will likely get worse as we go.

[atp1916]

It was a bit sad, people were selling off their rigs and such.

Either way, let's see what you folks can do! 

[jasinlee]

Ok, I got the go ahead, Hauntingshade suggested I distribute the miners initially.

[jasinlee]

I was asked to make this for a visual representation by one of the trusted hosts. Enjoy.

[milkmanHH]

Will these asics be able to vanitygen addresses?