Probably not far from the truth. For a workload like highly optimized image processing code, the simpler ARM instruction set does less work per instruction than modern x86. For a similarly-clocked M1 to beat out an i7, there's some amount of co-processor in there accelerating things. Either there's a fixed set of functions that M1 is faster at (like an old fixed-pipeline GPU), or the co-processor is closer to being a CISC processor with really deep sleep states.
Not that an i7 "coprocessor" would be a bad thing. We already down clock cores to save power; why not have a blend of i7 cores that turn all the way off until needed and some power-sipping ARM cores to do email.
If the figure in this article [techrepublic.com] is to be believed, the M1 has a lot real estate dedicated to accelerating a lot of different functions. I don't know how literally to take the figure as to the proportions of things but even just the number of different dedicated functions is quite large.
Where are the calculations that go with a calculated risk?
Unfortunately... (Score:5, Funny)
Re:Unfortunately... (Score:2)
Probably not far from the truth. For a workload like highly optimized image processing code, the simpler ARM instruction set does less work per instruction than modern x86. For a similarly-clocked M1 to beat out an i7, there's some amount of co-processor in there accelerating things. Either there's a fixed set of functions that M1 is faster at (like an old fixed-pipeline GPU), or the co-processor is closer to being a CISC processor with really deep sleep states.
Not that an i7 "coprocessor" would be a bad thing. We already down clock cores to save power; why not have a blend of i7 cores that turn all the way off until needed and some power-sipping ARM cores to do email.
Re: (Score:2)