I’m very new to bitcoin and I’m attempting to grasp a bit about mining concept. I don’t dispute the conference of utilizing ASIC primarily based computation for bitcoin mining, however quite I am interested in why they outperform co-processors like this for mining. Possibly a nuanced dialogue of some of the varied trade-offs could be the best and clearest option to perceive issues. Let me break it down like this:
Parallelism: Co-processors are largely used for process parallel code (non-vectorized code) whereas GPUs are adept at information parallel (vectorized) code. Two questions come to thoughts:
- Are all mining codes in vectorized languages?
- Can there ever be a profit to mine in process parallel code? It looks like with bitcoin mining, there’s only one process: crunching numbers. Am I oversimplifying it although? Maybe with process parallelism one might run different algorithms to help with the quantity crunching (similar to compression, flush to zero, ect).
Calculations: Talking of flush to zero, GPU’s are likely to carry out higher than co-processors with floating level calculations. Nonetheless, co-processors carry out higher in logical in addition to arithmetic calculations.
- Is floating level calculation extra vital than logical/arithmetic in bitcoin mining?
Latency:
Since GPUs run in sync, the entire dataset needs to be transferred at the start and finish of every process, however co-processors can transfer all the info from RAM to Host RAM in a fraction of a second. I am unsure if that is true of ASIC processors, perhaps it is completely different.
- Would the difficulty of latency ever warrant the usage of co-processors for mining, or is latency not an enormous difficulty?
Major Query May there ever be a pure co-processor mining machine? What a couple of hybrid system utilizing an array of co-processors and GPUs for one of the best of each worlds? Or would GPUs beat co-processors each time? Why or why not?
You might method this query from any angle as per the content material above. The reply would not need to reply all questions, I believed I might simply embrace a number of interrelated questions for robustness/comparability functions.
I wish to stipulate some assumptions to ensure the reply is usually from a theoretical standpoint. I wish to make the logistics of all of it largely out of scope.
Assumptions
- Energy consumption is out of scope
- Unit value of GPUs and co-processors is out of scope (clearly
co-processors are ungodly costly) - Value of electrical energy is out of scope
