I remain unconvinced that this is some big paradigm shift, and that the instruction set itself is mostly irrelevant for battery life and performance per watt.
Yes, Apple achieved a big jump with its first M1 at delivering some pretty amazing performance per watt, compared to contemporary chips from Intel.
But a closer look has shown that each successive generation of M-series Apple Silicon has been chasing higher performance at the cost of energy efficiency. Which is fine, but shrinks the gap.
And then, if you look at AMD’s low power x86_64 CPUs for laptops, you’ll see that they’re also able to deliver significant power savings compared to Intel. Comparing like for like, in terms of TSMC node, you see that AMD performance per watt seems to be in line with Apple’s. It’s just that Apple’s comparative advantage in business/legal strategy (not engineering) has them locking up TSMC capacity earlier.
Finally, a comparison of Apple’s mobile ARM SoCs to other manufacturers’ mobile ARM SoCs (including Qualcomm and Samsung) shows that Apple has a significant performance/efficiency lead over even other ARM chips.
So it’s probably not the instruction set. It’s just the engineering of the chips themselves, boosted by Apple’s business/logistics strategies getting their products to market first.
I remain unconvinced that this is some big paradigm shift, and that the instruction set itself is mostly irrelevant for battery life and performance per watt.
Yes, Apple achieved a big jump with its first M1 at delivering some pretty amazing performance per watt, compared to contemporary chips from Intel.
But a closer look has shown that each successive generation of M-series Apple Silicon has been chasing higher performance at the cost of energy efficiency. Which is fine, but shrinks the gap.
And then, if you look at AMD’s low power x86_64 CPUs for laptops, you’ll see that they’re also able to deliver significant power savings compared to Intel. Comparing like for like, in terms of TSMC node, you see that AMD performance per watt seems to be in line with Apple’s. It’s just that Apple’s comparative advantage in business/legal strategy (not engineering) has them locking up TSMC capacity earlier.
Finally, a comparison of Apple’s mobile ARM SoCs to other manufacturers’ mobile ARM SoCs (including Qualcomm and Samsung) shows that Apple has a significant performance/efficiency lead over even other ARM chips.
So it’s probably not the instruction set. It’s just the engineering of the chips themselves, boosted by Apple’s business/logistics strategies getting their products to market first.