In a website section called Getting Started with TI Processors, TI has a good general overview piece, Getting to Know Processor Architectures. This is at a pretty high level, but it’s solid, and helps explain the different approaches. The drill down information, not surprisingly, covers TI-specific processors, but overall the information may help you decide which processing option is the right on for your application. As the TI article says:
Selecting the right processor involves more than just estimating performance requirements.
The topics covered:
Microcontrollers (MCUs) are programmable, small in size, and low cost, and TI considers them “ideal for motor control, capacitive touch, measurement, sensing, lighting and energy.” Using an MCU has time-to-market and flexibility advantages over hard-coded approaches.
ARM and general purpose processors (GPP) are also programmable and flexible. ARM in itself is not a processor, but a core that vendors like TI use as building blocks to create a complete chip-level processor solution around. When combined with DSPs or microcontrollers, they’re used for applications like video and multimedia, as well as industrial applications.
Historically, we’re partial to Digital Signal Processors. Programmable, flexible, efficient. As TI says: “DSPs are ideal to apply embedded analytics to end-equipments…. DSPs are a great choice when low power, high performance, feature flexibility and time to market are key concerns.” Yep!
TI’s take on Application-Specific Integrated Circuits (ASICs) is that, given the cost of development and their lack of flexibility – not a great attribute when markets and technology change so rapidly – is that they need to be very carefully thought out. The big advantage of sinking the investment dollars into an ASIC development is the low recurring cost of the IC once it’s complete and verified. Take on ASIC development only when there is no other way to get the job done, or the extremely high volume of the end product warrants the small footprint or low recurring price tag.
Field Programmable Gate Array (FPGAs) “provide performance close to ASICs without the delays and costs associated with respinning an ASIC.” However, TI views them as difficult to work with. We don’t see them in quite the same way, but concur with TI that, “for certain applications, FPGA can be used to complement the DSP.” We offer an optional FPGA on many of our SoMs.
Anyway, I’ve pulled out some of the highlights, but the full TI article is worth a read, especially if you’re new to making decisions about processors.
I did want to point out one thing that’s not mentioned, and that’s microprocessor units (MPUs). The primary difference between MCU and MPU is that the MCU has on-chip memory, and the MPU requires off-chip memory in the board design. MPU’s typically have more horsepower. In their article, what TI is calling a GPP is generally classified as an MPU.