Re: [AXIOM] AxiomMemHash, Schnorr Sigs Implemented, APOS 3.0, AXH 2.0 Proposed
This one is not being used as a miner at the moment. 
I only made it, because of the claims that this algo is GPU resistant. Well... it is not and probably no matter what kind of algo they come out with, GPU will always have slight advantage over CPU. Simply because, it is better in parallel computations. And if you make single hash calculation very expensive so that it has hard times running on GPU, then even CPUs will struggle, not to mention older PCs with slower CPUs - that would render wallet not being useful anymore.
GTX 980 Ti uses 200W (at wall) to produce ~1.3 kH/s. CPU using same power produces ~0.4 kH/s. GPU is therefore approx 3 times more efficient.
I only made it, because of the claims that this algo is GPU resistant. Well... it is not and probably no matter what kind of algo they come out with, GPU will always have slight advantage over CPU. Simply because, it is better in parallel computations. And if you make single hash calculation very expensive so that it has hard times running on GPU, then even CPUs will struggle, not to mention older PCs with slower CPUs - that would render wallet not being useful anymore.
GTX 980 Ti uses 200W (at wall) to produce ~1.3 kH/s. CPU using same power produces ~0.4 kH/s. GPU is therefore approx 3 times more efficient.
GTX has much overhead than 200W, costs about $500 each card and creates a ton of kelvin. You are thinking a little small. When you have a blade server, you can produce 600 h/s at 150W per blade at $450 each, very little heat, plus my devices don't lose value as quick as a GPU. You haven't quite caught up to me until you hit about 2500 h/s according to all the factors.
Long and the short, you should buy blades rather than GPUs. GPU miners fade fast, X11 is over.