4 ms·

GPU mining is still not as decentralized, as basically only high-end PC gamers (with their own default cards) and the people specifically buying GPUs for mining

by salient 13y ago

GPU mining is still not as decentralized, as basically only high-end PC gamers (with their own default cards) and the people specifically buying GPUs for mining will have a shot at it. CPU-only mining would be a lot more decentralized, if such a thing is even possible.

lukifer 13y ago
ProtoShares (and a few other coins, I believe) focus as much on being memory-hard as CPU-hard. While CPU/GPU can be "cheated" by designing ASICs, there's no such thing as application-specific memory, making it a better candidate for an honest signal. [1] [1] http://en.wikipedia.org/wiki/Signalling_theory http://en.wikipedia.org/wiki/Signalling_theory
- erikpukinskis 13y ago
  Won't that just lead to people making machines with giant racks of RAM, causing the exact same difficulty race as we're seeing with the giant racks of ASCIs in Bitcoin? And even if that's not the optimization (I'm not an expert on ProtoShares or memory-constrained algorithms), I am highly skeptical of anyone claiming a computation problem is optimally solved by a laptop, and cannot be optimized with custom engineering.
  - woah 13y ago
    Giant racks of RAM could be useful. Once someone's rig is no longer profitable for mining, they could turn around and open a MongoDB hosting service.
  - lukifer 13y ago
    It's about cost versus yield. Custom-designed ASIC chipsets are orders of magnitude more efficient at their intended task compared to general-purpose CPUs; whereas it's impossible to buy or make "special" RAM, and cheap RAM is something everybody wants for everything anyway.
acangiano 13y ago
CPU-mining however is also more profitable for malicious parties (e.g., botnets).
- whyenot 13y ago
  That may be true, but there isn't any convincing evidence that botnets are targeting (for example Primecoin and Protoshares), at least I am not aware of any, and I follow this area fairly closely.
tedunangst 13y ago
One of the difficulty parameters for scrypt is memory. Set it to 16GB and you'll keep GPUs out of the game for a while.

by salient 13y ago

lukifer 13y ago

ProtoShares (and a few other coins, I believe) focus as much on being memory-hard as CPU-hard. While CPU/GPU can be "cheated" by designing ASICs, there's no such thing as application-specific memory, making it a better candidate for an honest signal. [1] [1] http://en.wikipedia.org/wiki/Signalling_theory http://en.wikipedia.org/wiki/Signalling_theory

erikpukinskis 13y ago
Won't that just lead to people making machines with giant racks of RAM, causing the exact same difficulty race as we're seeing with the giant racks of ASCIs in Bitcoin? And even if that's not the optimization (I'm not an expert on ProtoShares or memory-constrained algorithms), I am highly skeptical of anyone claiming a computation problem is optimally solved by a laptop, and cannot be optimized with custom engineering.
- woah 13y ago
  Giant racks of RAM could be useful. Once someone's rig is no longer profitable for mining, they could turn around and open a MongoDB hosting service.
- lukifer 13y ago
  It's about cost versus yield. Custom-designed ASIC chipsets are orders of magnitude more efficient at their intended task compared to general-purpose CPUs; whereas it's impossible to buy or make "special" RAM, and cheap RAM is something everybody wants for everything anyway.

acangiano 13y ago

CPU-mining however is also more profitable for malicious parties (e.g., botnets).

whyenot 13y ago
That may be true, but there isn't any convincing evidence that botnets are targeting (for example Primecoin and Protoshares), at least I am not aware of any, and I follow this area fairly closely.

tedunangst 13y ago

One of the difficulty parameters for scrypt is memory. Set it to 16GB and you'll keep GPUs out of the game for a while.