 |
March 26, 2014, 10:30:57 AM |
|
I'm not sure, for now making a dedicated sha3 coin mining asic isn't feasible, but given how this coin progresses as time goes by it certainly WILL be justified.
The way vertcoin or scrypt-chacha coins do it works, it'll keep asics away of those algorithms, since the n factor changes are fast, vertcoin not so much. I can see people making asics for it, as long as the developer keeps the current one year time frame. I don't think constant hard forks are good for the coin's image though, so I can see that as a problem in the future, given that vtc's asic resistance is it's main selling point, basically litecoin 2.0.
Myriad is ahead with the way it includes and welcomes all miners, and all types of devices. I'm specially fond of the groestl algo for older cards, and the skein and qubit one run great on newer hardware, whilst saving power.
A hard fork shouldn't be necessary for now, as long as it's more profitable for asic manufacturers to focus on scrypt and sha256 we should be safe. If it turns out to be needed, then I'd suggest incorporating a high N factor algo in the mix, maybe to replace qubit or skein and maybe implement X11 instead of the other? AFAIK drk's algo is asic proof too, but not sure how much ram it uses, it can be adapted to asics eventually as well.
N factor 12 and up scrypt-chacha equalizes old hardware with newer hardware, gpu ram and system ram being the limiting factor, in relation to the amount of shaders.
Since most cards are equipped with 1 to 4gb vram buffers, this causes lower end cards with more ram to perform equally as well to higher end cards with the same amount of ram and more shaders*
(the algorithm uses big amounts of ram per thread, and to saturate all shaders with work and max out hash rate you need more ram than the gpu has, at least with 90% of the cards out there, forcing you to use lookup gap to diminish the thread's ram usage, which introduces a performance penalty, which can be softened by a high core overclock)
|
