From Jareso:@bubus Update info regarding my effort of making fast Yespower 1.0 miner for Nvidia GPUs.
In recent weeks, as Yenten transition to new algorithm gets closer, I worked hard on my experimental prototype of Yespower 1.0 Nvidia GPU miner for all its N, R variants programing all sorts of optimizations to my Yespower 1.0 CUDA kernels. Several times I even rewriten my Yespower 1.0 CUDA kernels, redesigning algorithmic approach and even with all of this effort, with tons of time and optimization work I spend on it, the fastest where it managed to reach is 550 H/s (550 hashes) on one single 1080 Ti at Yespower 1.0 N 2048, R 32. Similar for N 4096, R 16.
It's waaaaaay tooooo slooooow.
So, I can basically say that if Yenten wants to be CPU dominant coin than Yespower 1.0 is indeed a great choice to stick to!
It's very unlikely that GPUs will be able to overtake Yespower 1.0.
Optimization techniques, tricks and CUDA speed-ups that work great for Yescrypt and other mining algorithms are basically useless for Yespower 1.0.
Despise being skeptical now for fast Yespower 1.0 on GPU, I still have more ideas in my mind about more approaches how to proceed to Yespower 1.0 on GPU which I haven’t tried so far.
Of course, as I said the last time, it is always very hard to predict hashrate of certain algorithm and various optimizations in advance – without actually implementing and running them.
Those will again require tons of programing effort and time with questionable outcome, but anyway, I actually like things like that.
When things don’t go as I expected, it gives me more strength to the challenge to find some way how I will finally twist neck of that Yespower 1.0 on GPU, one way or another. :grimacing: :smile: :wink:
But to be objective, the outcome is that Yespower 1.0 seems to be so bullet-proof algorithm that probably not even in years from now it will be dominated by GPUs, not even by the highest-end future GPUs.
We express our gratitude for the great work!!!
yes this is incredibly valuable work.
our community has risen to the occassion over the last year to right the ship and set Yenten up for a successful future. in the beginning, i had my doubts whether development would progress without conan revealing himself. now, i understand that there are many talented community members who are willing to contribute time and effort deapite seeing zero compensation in official dev fees.
although we had already achieved community consensus, Jareso’s work goes a long way to solidify my support for the algo change.