You all are familiar with the concept of a microprocessor in your computer. Intel and AMD collectively own the market for that part in personal computers. I suspect however that less is known about the computing part in embedded devices such as phones and consumer electronic devices.
On the PC/Mac where many applications have been written from years past, maintaining backward compatibility with Intel instruction set (which AMD has cloned) has locked the market into those two companies as providers. In embedded space however, such legacy did not exist and companies were free to pick the instruction set (the language of the computer) of their choice. One such choice was the one defined by a UK company called ARM.
ARM’s claim to fame was power efficiency over absolute performance. In the PC world, that equation is somewhat reversed in the chase for ever increasing CPU horsepower to convince customers to upgrade. In contrast, devices such as phones that run on battery power need to run efficiently. Needless to say, the rise and popularity of product such as smartphones has led to meteoric rise of ARM implementations.
ARM is not a microprocessor company however. Instead, it creates designs that other companies license and then create real parts around. Think of them as the architects for a building and you, the company licensing it, as the general contractor. Designing microprocessors and tools to program it is difficult and by licensing all of that know-how, a company can have its own custom microprocessor with far less resources and faster time to market.
By buying the CPU “core” from ARM, the licensee can then focus its efforts around customizing the part around his own need. They can integrate other peripherals on the chip such as memory, LCD controller, Graphics controller, etc. To the extent such a chip encapsulates most of the functionality of a device, it is called a System on a Chip or SoC for short.
SoCs are the heart of just about any “smart” device out there. Your Blu-ray player has one as does your iPhone or Samsung/Motorola phone. They reduce cost of the system, and importantly shrink the amount of space it takes. The latter is extremely important in phones and tablets where space is a huge premium. Everyone wants smaller devices and SoCs are the critical enabling technology for that.
Back to licensing the microprocessor from ARM, there are two paths to a real chip. One is to take an already designed logic block and then work with an independent silicon company to produce it. The most well-known company in that industry is TSMC in Taiwan. TSMC has world-class chip manufacturing facility riving that of Intel and others at the bleeding edge. Most companies go this route. Apple uses Samsung for production of its silicon, likely because they can then embed their memories on chip. The Apple A6 processor is said to have 1 Gigabytes of memory on board the CPU package.
An alternate path exists where the company uses the instruction set from ARM but designs its own “core.” In this case the company has to design the logic and in doing so, take on more of the risks and rewards. The risk is that it doesn’t fully verify the chip and issues come up in execution. The reward is that they can get a more optimized microprocessor/SoC design. Apple’s A6 processor used in iPhone 5 is an example of such a custom implementation whereas the earlier designs used licensed cores. Android smartphone manufacturers use silicon from other companies which provide that role such as Qualcomm. As with Apple, Qualcomm has both custom and licensed ARM cores in its portfolio.
There was a time that such designs heavily lagged the PC in speed and performance. Today, the clock speed (how fast the wheel turns in a microprocessor executing instructions) has cross the Gigahertz barrier and is moving up from there. Likewise, we are already in dual core domain, moving to quad core. Performance is probably 2-10 times slower than the PC still but given the lighter workloads, it is not an impediment to good performance.
On the PC/Mac where many applications have been written from years past, maintaining backward compatibility with Intel instruction set (which AMD has cloned) has locked the market into those two companies as providers. In embedded space however, such legacy did not exist and companies were free to pick the instruction set (the language of the computer) of their choice. One such choice was the one defined by a UK company called ARM.
ARM’s claim to fame was power efficiency over absolute performance. In the PC world, that equation is somewhat reversed in the chase for ever increasing CPU horsepower to convince customers to upgrade. In contrast, devices such as phones that run on battery power need to run efficiently. Needless to say, the rise and popularity of product such as smartphones has led to meteoric rise of ARM implementations.
ARM is not a microprocessor company however. Instead, it creates designs that other companies license and then create real parts around. Think of them as the architects for a building and you, the company licensing it, as the general contractor. Designing microprocessors and tools to program it is difficult and by licensing all of that know-how, a company can have its own custom microprocessor with far less resources and faster time to market.
By buying the CPU “core” from ARM, the licensee can then focus its efforts around customizing the part around his own need. They can integrate other peripherals on the chip such as memory, LCD controller, Graphics controller, etc. To the extent such a chip encapsulates most of the functionality of a device, it is called a System on a Chip or SoC for short.
SoCs are the heart of just about any “smart” device out there. Your Blu-ray player has one as does your iPhone or Samsung/Motorola phone. They reduce cost of the system, and importantly shrink the amount of space it takes. The latter is extremely important in phones and tablets where space is a huge premium. Everyone wants smaller devices and SoCs are the critical enabling technology for that.
Back to licensing the microprocessor from ARM, there are two paths to a real chip. One is to take an already designed logic block and then work with an independent silicon company to produce it. The most well-known company in that industry is TSMC in Taiwan. TSMC has world-class chip manufacturing facility riving that of Intel and others at the bleeding edge. Most companies go this route. Apple uses Samsung for production of its silicon, likely because they can then embed their memories on chip. The Apple A6 processor is said to have 1 Gigabytes of memory on board the CPU package.
An alternate path exists where the company uses the instruction set from ARM but designs its own “core.” In this case the company has to design the logic and in doing so, take on more of the risks and rewards. The risk is that it doesn’t fully verify the chip and issues come up in execution. The reward is that they can get a more optimized microprocessor/SoC design. Apple’s A6 processor used in iPhone 5 is an example of such a custom implementation whereas the earlier designs used licensed cores. Android smartphone manufacturers use silicon from other companies which provide that role such as Qualcomm. As with Apple, Qualcomm has both custom and licensed ARM cores in its portfolio.
There was a time that such designs heavily lagged the PC in speed and performance. Today, the clock speed (how fast the wheel turns in a microprocessor executing instructions) has cross the Gigahertz barrier and is moving up from there. Likewise, we are already in dual core domain, moving to quad core. Performance is probably 2-10 times slower than the PC still but given the lighter workloads, it is not an impediment to good performance.
