Steve Wozniak in 1976 designed the Apple 1 while working at HP during the daytime, and he used standard parts to keep costs low, like:
- 6502 CPU from MOS Technology
- 8K of DRAM
- TTL logic for driving video and random logic
- PROM to hold the BASIC language and primitive OS
Oddly enough, HP management didn’t see a future in the nascent, personal computer market, so they passed on making the Apple 1 an HP product.
The Apple II followed the Apple 1,and continued to use off the shelf parts. Starting with the Apple IIc and Macintosh computers there was a custom chip designed and dubbed IWM – Integrated Woz Machine, that controlled the floppy disk-drive. Apple could keep the total computer part count and costs lower by designing their own custom chips and having them fabricated by foundries like VLSI Technology.
By the mid 1980’s the Apple engineers had designed an ASIC to integrate all of the Apple II logic board into a single chip called the Mega II. Apple continued to design their own custom audio, graphics, SCSI controller and other glue components. Commodity parts were purchased, like the: CPU, RAM, ROM, power control. Gate array or standard cell technology was used to keep the costs of these ASIC chips low, in contrast to full-custom design which is typically reserved for GPU, CPU and memory chip designs.
Processors
Apple has used three separate CPU vendors over the years to power their desktop and server products:
- Motorola 68K from 1983 to 1990, Lisa and Macintosh products. Motorola 68020, their first 32-bit processor in 1987.
- PowerPC from 1994 to 2005.
- Intel (2006-present) – desktop and servers run Mac OS X.
PDA
Handheld devices like Personal Digital Assistants (PDA) required long battery life to be commercially viable, and using processors designed for the desktop was not a viable power strategy, so Apple sought out a provider of low-power processors. In 1990 ARM was funded as a spin-off from Apple, Acorn and VLSI, and an ARM6 chip was designed for Apple’s first tablet, the Newton in 1992. The Newton failed because it promised hand-writing recognition which never quite worked, so less-sophisticated PDAs like the Palm Pilot ruled that market because their graffiti-writing worked well and was fast enough for minimal note-taking.
ARM
Apple really wanted to craft their own processor for the handheld market instead of using CPUs from Motorola or IBM in order to meet power and cost goals, so in 2008 they bought a chip-design company called P.A. Semi. Before that acquisition P.A. Semi was designing processors for the 64-bit Power Architecture, an alliance of Apple, IBM and Motorola/Freescale. Now Apple could create full-custom chips to rival any CPU or GPU company and gain control of their own destiny.
The first iPhone in 2007 used an ARM-based CPU and GPU combination called the APL0098, fabricated at Samsung on a 90nm process. In 2008 an integrated CPU and GPU called the APL0278 powered the iPod Touch (2nd generation), manufactured at Samsung in 65nm technology. For the iPhone 3Gs the CPU/GPU was APL0298, and it used the 65nm node from Samsung.
A4
The first iPad used an A4 chip in April 2010, designed by Apple and fabricated by Samsung in 45nm. The A4 was also used in the iPad Touch (4th generation), iPhone 4 and Apple TV (2nd gen) devices. Unit volumes are high in most of Apple’s consumer products, so then can amortize the development costs of A-series processors over multiple products, always trying to keep a step ahead of any competition. A novel thing about the A4 is how it uses Package on Package (PoP) technology to add 256 MB of DDR SDRAM inside, which helps to support the high graphics bandwidth requirements.
Intrinsity
Apple acquired Intrinsity in 2010, a processor design company from Austin, Texas with about 100 engineers that had experience in low-power design using dynamic logic. This methodology would help Apple keep the A-series processors very low power compared to any processor off the shelf.
A5
Unit volumes for both the iPhone and iPad products were ramping, and they enabled Apple to stay ahead of competitors in terms of user experience like battery life, and responsiveness. The A5 CPU/GPU was designed by Apple, then fabricated at Samsung in both 45nm (2011) and 32nm (2012) versions, and used in the iPad 2, iPhone 4s, Apple TV (3rd generation), iPod Touch (5th generation) and iPad Mini products.
A5X
Apple continued to obsolete their own products by developing replacements at a pace of about one new generation per year, all with combined CPU and GPU features that were custom-designed. The A5X introduced in the iPad 3 was a quad core, and debuted in 2012, and fabricated at Samsung in 45nm. I purchased the iPad 3 new and still use it today, along with a Logitech Ultrathin keyboard, a nice combination for us touch-typists.
A6
The A6 chip was designed for the iPhone 5, where Apple bought an ARM license, yet designed their own processor instead of using a reference design from ARM. This was a late 2012 announcement and the A6 chips were manufactured by Samsung in a 32nm HKMG (High K Metal Gate) process.
A6X
Replacing the A6 just one month later in 2012 was the A6X used in the iPad 4 and fabricated at Samsung in their HKMG 32nm process.
A7
In September 2013 Apple announced their first 64-bit processor as the A7, used in the iPhone 5s and iPad Air, and built in a 28nm HKMG process by Samsung.
Apple A7 die photo, over 1 billion transistors, source: Chipworks
Summary
Apple has come a long way in designing their own custom chips for use in both desktop and mobile devices, spanning from 100,00 transistor designs to over 1,000,000,000 transistors. The trend is clear that Apple will continue to design their own CPU/GPU A-series processors for the mobile market, and the next big change for the desktop could be a migration away from Intel-powered CPUs to an all-Apple design. It all comes down to a decision of economics, delivery, compatibility, control, and satisfying each market. Stay tuned to SemiWiki to hear about which foundry will continue serving Apple beyond the A7.
Next Generation of Systems Design at Siemens