bsimpsen

In the last few weeks, a significant fraction of my spam calls (2-5/day) have been coming from the UK. I suspect they anticipated the legislation.

I began my four decade career in electrical engineering as a QA/QC engineer for a large company making building automation/safety systems. Reliability was tantamount. While analyzing field failure reports, I discovered, unsurprisingly, that connectors were the highest failure rate items. The military reliability handbook (MIL-217) warned of this, so it was not a surprise. The fascinating aspect of the field failure data I received was that, although connectors of almost any kind truly were more prone to failure than most other things, socketed parts of systems (chips in particular) were themselves "failing" at above expected rates. Deeper analysis revealed that, because socketed things are easy/cheap to replace, they would be routinely replaced, and because the replacement wasn't done under controlled factory conditions, connection reliability decreased.

There is ample historical data to support designing systems to have as few mechanical connections as possible, and that's precisely what Apple does. While repair of such systems may be more difficult, the reduction in overall failure rate more than makes up for it. The truth of this is reflected in Apple's consistently high user satisfaction scores.

I was hoping for details of how Apple integrated the various semiconductor companies they have bought over the years. To me that is the killer part of the story.

Absolutely.

PA Semi brought significant CPU architecture experience to Apple. Intrinsity brought significant optimization expertise. I recall that Apple also bought up some GPU expertise in Florida in addition to the staff they obtained from Imagination Technologies. They acquired substantial Flash memory storage optimization expertise in Anobit, biometric ID capability in AuthenTec, multi-lens camera tech in LinX and 3D optical sensing in PrimeSense. And now they have $1Billion worth of Intel's 5G modem expertise and a significant portfolio of power management IP from Dialog.

I don't think there is another company on Earth that has this much wood to put behind their design arrow. And, unlike Qualcomm, Apple knows exactly where it wants to shoot that arrow.

netmage said:

They took the only course available.

No, while Intrinsity and Hummingbird might have been off the table, pursuing a course of designing their own successor was available to them, through purchase of another chip designer or starting from scratch, one they eventually attempted with Exynos, though that seems like it might be winding down now as a failed attempt. But who knows where they might be if they had started a few years earlier to found their own design path. 

Samsung did start a few years earlier, in concert with Intrinsity. Apple pulled the plug on that, and Samsung was forced to change course. That course was eventually Exynos.

AppleInsider said:

Samsung also used the Hummingbird core and PowerVR GPU in its chip, which was later branded as "Exynos 3." But rather than seeking to relentlessly advance its custom chip design technology in the pattern of Apple, Samsung initially took the more comfortable and affordable route of relying on ARM to deliver its Cortex-A CPU and Mali GPU designs. That didn't work out well.

I don't think you got the history quite right here. The Hummingbird core was a joint project between Intrinsity and Samsung, with the bulk of the design credit going to Intensity and Samsung acting as the foundry. The A4 was based on that work, and debuted in Apple's first iPad in April 3 of 2010. On April 27, 2010, Apple acquired Intensity, depriving Samsung of the ability to advance the Hummingbird architecture. It was not so much that Samsung took the more comfortable course of relying on ARM. They took the only course available.