I haven't looked at CPU architectural comparisons, but Apple must have a major benefit in owning the whole stack: from the lowest hardware level to the OS to many of the applications. And with the right focus (which they almost certainly have) the ability to optimize the system as a whole instead of each level taking care of just their own section.
If the x86 translation team can make their case that a specific CPU feature will be a significant enough performance benefit, chances are that a discussion within the company will have less friction that a discussion between, say, a Microsoft SW team and a Qualcomm or ARM CPU design team.
And while Microsoft may care about power consumption and power management, I can't image that they care as much as the Apple MacBook team. (My POS high-end Dell laptop is still a power management disaster. It would fail any internal go-to-market review at Apple.)
But there's also the simple fact that the Apple silicon team is very good, and they've been cranking out best in class silicon IPs for more than a decade. A lot of IPs on an SOC are developed incrementally, with smaller improvements from one generation to the next accumulating into major advancements over the years. Even if the competition builds an equally capable silicon team, they may not have the solid foundation to build on and it can take generations to match what the competition already had.
It's also true that for Apple, the silicon is not a product that's being sold on its own. I'm sure they care about silicon area (and thus cost), but they can once again put this cost in the larger context of the full laptop. For Qualcomm, increasing a cache by 50% is a much bigger hit on their gross margins than for Apple. If that cache increase will result in significantly better laptop characteristics, Apple may decide to go for it while Qualcomm may not.
For Dell, the customer is corporate Information Technology departments and accountants. Most of their end-user kit is minimum viable brick for business use: if it isn't in the box, it can't break. I find their engineering impressive.
Apple tells corporate IT to get out of the way of their customers, the users. Thus repairing their stuff is a nightmare.
(My laptop is an M1 MacBook Air. My servers are off-lease salvage Dell or Intel OEM hardware.)
It’s not just TSMC - they also have a very good CPU team giving AMD heartburn even when you compare the chips on the same TSMC process, and they have the part which is hardest to match: they control the OS, compilers, libraries, and key apps. That allows them to ensure that hardware acceleration is actually used, tailor chips for the power and usage patterns their devices need, and avoid the conflicts of interest inherent to multi-vendor systems because they share responsibility for a great device and profits aren’t unevenly distributed. That’s not perfect and they’re certainly in a bad place if they misjudge but it’s worked so far.
Apple's core "Silicon" team was the Centaur Technologies chip design group. At the time, they had brought to market a PowerPC CPU chip that was a breakthrough in performance per watt -- for 32-bit PowerPC. Which had already lost to Intel for mainstream applications.
At Apple, Centaur joined engineers who had helped to develop early ARM RISC processors.
While I know that the talent pool migrates, engineers have moved on, Apple has retained deep institutional knowledge of how to build systems that work really well. And developed teams worldwide to deliver manufacturing at scale.
By no means are they the only company to do so. Vast resources in competition. But please understand that Apple isn't new to full-stack hardware design.
If the x86 translation team can make their case that a specific CPU feature will be a significant enough performance benefit, chances are that a discussion within the company will have less friction that a discussion between, say, a Microsoft SW team and a Qualcomm or ARM CPU design team.
And while Microsoft may care about power consumption and power management, I can't image that they care as much as the Apple MacBook team. (My POS high-end Dell laptop is still a power management disaster. It would fail any internal go-to-market review at Apple.)
But there's also the simple fact that the Apple silicon team is very good, and they've been cranking out best in class silicon IPs for more than a decade. A lot of IPs on an SOC are developed incrementally, with smaller improvements from one generation to the next accumulating into major advancements over the years. Even if the competition builds an equally capable silicon team, they may not have the solid foundation to build on and it can take generations to match what the competition already had.
It's also true that for Apple, the silicon is not a product that's being sold on its own. I'm sure they care about silicon area (and thus cost), but they can once again put this cost in the larger context of the full laptop. For Qualcomm, increasing a cache by 50% is a much bigger hit on their gross margins than for Apple. If that cache increase will result in significantly better laptop characteristics, Apple may decide to go for it while Qualcomm may not.