Hi there!

Guys, may someone tell me what speed on Kangaroo and full priv <-> address brute (aka bitcrack implementation) could expect for RX 580 8 Gb, GTX 1070 ti, RTX 2070 - and your justification for your assumptions.

We know there are several implementations algorithms for different platforms and yet i've seen different speed numbers and so wondered "why?":

GTX 1070 ti has 2432 CUDA cores

RTX 2070 has  2304 CUDA cores

RX 580 has 2048 stream processors

And all have different architecture - sure i understands that (even Turing dramatically differs from Pascal in sense of "instructions" set or pipe or how to call this differences)

======================

Anyway, from the first glance i could expect close speeds from these three cards and so i am asking here what really could be achieved with these cards and the reason for limit.

P.S. Have been tried Kangaroo on my RTX 2070 (~1000 MKey/s) - found that memory overclock plays significant role that means that memory in the case of RTX 2070 (and probably my config) was a limiting factor (CUDA cores probably but not surely could provide more computational power == more speed)

P.S.S. I am on the south of Ukraine under occupation mostly without money and having bad mood - so i've returned to this problem just to smooth my mood a bit.


(https://github.com/HomelessPhD/BTC32)