The mining hardware gets a base of work that it can send off unique starting work to trillions of hashing units. Unless the hardware is a broken design there is zero chance of their work overlapping. Speed is irrelevant, every single hash unit operates on its own independently.
Hi -ck
Thanks for the response.
Arguably, isnt it the miner software that issues starting nonce to an ASIC, not hardware.
So while a pool may issue a base of work to the miner software, the miner software must surely determine if that base of work is followed and what starting nonce to issue to an ASIC.
Which means the miner software must track what nonce's it has issued from the base of work that would be sent to an ASIC in order to prevent overlap or rehasing from a previously used nonce.
Thanks for your help, just trying to make sense of the flow of work.
I've dug through the code of CGMiner to try and understand how the work is issued, but it is a very complex piece of software with little in the way of code level documentation making it quite hard to understand the program flow.