Topic: YACoin - Why have GPU miners? (Read 908 times)
Interesting. Disappointing but interesting.
From what we have seen, GPU mining YAC will always be orders of magnitude faster than CPU mining.
If this is true, then YACoin is pointless. It's main objective was to offer reasonable CPU mining. If folks want a coin that has failed in this objective already, there is always Litecoin.
My *strong* feeling for this design is the following :
The original dev (who disappeared a few days after launch)
- either had a private GPU miner ready at launch
- had access to a lot of CPU, but not much GPU. The changes in the hash function were designed as a deterrent for all the people having access to GPU farms.
Either way, he probably made a lot, dumped and disappeared.
Because the windows miner was initially crippled, the second option is probably the right one ... To further boost is profit, the easiest way was to silently reduce initial mining ability of many interested people. Should he had a GPU miner at launch, he would probably not have gone that way.
It was obvious from the start that yacoin would be GPU minable ... sha3/chacha were a surprise tough, only the N factor had been presented before launch ... (and sha3 is not trivial to implement in opencl if you are not already familiar with opencl)
Also, look at the current distribution of wealth ... more than 2/3 of all yacoins are kept in less than 50 addresses.
From what we have seen, GPU mining YAC will always be orders of magnitude faster than CPU mining.
If this is true, then YACoin is pointless. It's main objective was to offer reasonable CPU mining. If folks want a coin that has failed in this objective already, there is always Litecoin.
Hm, how does that work? GPUs are around the 100Kh/s zone at current N for YaCoin, how will they get to 1Mh/s at the highest N?
Nfactor <> NWe are at NFactor=10 and N=2048 for Yacoin at the moment with Ybcoin NFactor=4 and N=32.
Derp.
Why would they clone YaCoin to start at 32... that's dumb now.
Why would they clone YaCoin to start at 32... that's dumb now.
Hm, how does that work? GPUs are around the 100Kh/s zone at current N for YaCoin, how will they get to 1Mh/s at the highest N?
N = 32 is the smallest N for Yacoin, not the highest. N = 32 corresponds to Nfactor = 4.
Hm, how does that work? GPUs are around the 100Kh/s zone at current N for YaCoin, how will they get to 1Mh/s at the highest N?
orders of magnitude
I don't think there's any GPU able to get two orders of magnitude more hashes than a desktop CPU. My 6850 barely reaches one order of magnitude above my i5 2500k with the latest optimized kernel.
Well, it depends on the value of N. If you set N to 32, my latest kernel will do over 1 million hashes per second quite easily. That's what's happening at the Ybcoin camp. They launched with N = 32 a few days ago and their network hashrate seems to have reached 87 Mhash/s already. Most of the coins are probably going into the pockets of a few GPU miners.
orders of magnitude
I don't think there's any GPU able to get two orders of magnitude more hashes than a desktop CPU. My 6850 barely reaches one order of magnitude above my i5 2500k with the latest optimized kernel.
Oh, and scrypt-jane is only an implementation of scrypt that adds more mixing and hashing functions, it has nothing to do with Yacoin, and the fact that N is increased is given by a predefined function in Yacoin.
There is no reason to believe that CPU mining will be better at any N point.
From what we have seen, GPU mining YAC will always be orders of magnitude faster than CPU mining.
From what we have seen, GPU mining YAC will always be orders of magnitude faster than CPU mining.
If I understand correctly, YACoin uses Scrypt-Jane which increments the value of N over time, making GPU miners efficient at the start, but gradually less and less efficient over time, so that CPU mining eventually becomes the most efficient method of mining.
What's the philosophy behind that? Why bother with the GPU miners at all? Why not just start with a high N value so only CPU miners are efficient right from the outset?
What's the philosophy behind that? Why bother with the GPU miners at all? Why not just start with a high N value so only CPU miners are efficient right from the outset?
Jump to: