RE: Let me break this particular one down for you.17 Mar 2024 17:48
"Source, trust me Bro."
No offence I originally did but now I sort of don't.
Source:
https://github.com/skot/BM1397/blob/master/registers.md#bm1397-registers-map
Explanation. Each chip includes a,
https://github.com/skot/BM1397/blob/master/registers.md#chip-nonce-offset
https://github.com/skot/BM1397/blob/master/registers.md#hash-counting-number
https://github.com/skot/BM1397/blob/master/registers.md#ticket-mask
Therefore, at least at this level it is apparent that you can set a nonce offset or count starting point. I'm not sure about the counting number but the ticket mask is likely to be the number of leading zeroes required to satisfy the block and win the coins.
Given each chip on a hashboard is independently addressable and that is done by the firmware and the firmware is connected to the pool (or farm controller for that matter) I will state that it is possible, and you would be stupid not to do it, to at least assign starting points to every ASIC available from your resources to maximise your search range and avoid overlap. If you were to do this then such start points would only need updating when you add more resource.
That is the simple case.
However if we are to understand the meaning of QBT's PR then Method B supposedly guesses the most probable search ranges and as such they would have to adjust those starting points across the available resource. That becomes harder to do. However it is likely that they were forced to adopt an SaaS model, a central controller, to dole out those start points. In effect the SaaS solution is just another excuse for delays to excuse a poorly thought through methodology.
I'll go further, This talk of CGminer being hard coz it has 50,000 lines is, to me, also an excuse for delays or not knowing what they are doing. They should have immediately hacked the interface to the hash boards. It might be a 40 pin header and gives you direct control of the registers in the ASICs via their address lines and in particular the nonce-offset register. Then you program the internal controller to waggle the pins the way that you want to.
Instant Manufacturer Bypass.
Blaming CGminer for being bloated is also just being ignorant for purpose. CGminer covers a range of mining options. CPU, GPU and ASIC along with a number of Crypto Currencies that you can mine. It tries to be a Swiss Army Knife. Sure rig manufacturers might use it as a starting point but they will just compile the bits they need and, if necessary, hack them for communications. Stop moaning about 50K lines and RTFM.
Better still just bin it and hack the board communications interface. It's not Rocket Science.
Sign In
Follow the stocks
