X11 is rather straightforward, there are some caveats inside hash functions, but for well known hashes it would be just concatenation of published implementations.
For Quark algorithm one can implement 9 hashing stages, 3 of them will be paired - e.g. calculate Keccak and JH in parallel, then discard one of results. Or, to save die space, it's possible to implement 6 hashing cores and some dispatching and routing logic around them.
For Quark algorithm one can implement 9 hashing stages, 3 of them will be paired - e.g. calculate Keccak and JH in parallel, then discard one of results. Or, to save die space, it's possible to implement 6 hashing cores and some dispatching and routing logic around them.
Can you comment:
With sufficient funding, how long would it take to roll out usable ASICs based on a multi algo concept?
Do you think you can hold up to AMD and nVidia? They are not gonna miss that boat for very much longer.
Not only for crypto hashing, but for any kind of a custom instruction set for scientific applications.
They are not sleeping on trees.
Would a well meant ASIC be able to keep up with a 3000 stream GPU doing a custom program?
The noise of my GPU cards just said: NO!
you should tell your roaring gpu's to google search what "FPGA" means
edit:
@rumlazy i have been saying that too for a while and trying to see what people say.
and one guy on another site insisted it's not as simple as that actually.
i can't remember the tech details he posted but he explained that it would not be enough to tack on more memory to asics
because of how the asics work AND how they access and use memory i think.
I had originally said well memory is cheap so seems like an easy issue right ? add more memory then scrypt-n is now asic minable.
but he was rather convincing and sure as hell knew in remarkable detail how asics work under the hood specifically in ways i barely understand.
And the guy said no.. that won't be enough, simply adding more ram.
Either way i don't know but that guy was very convincing with his knowledge of how asics work.
i also am not sure tacking on memory still make it possible to jump from mining scrypt-n to normal scrypt either.
someone from the scrypt-n dev team(s) should address this stuff not me lol
i had thought the same but now i am not so convinced..