The microprocessor has clearly been a most significant driver of the electronics industry as we know it today. Its many advantages, including the flexibility offered by software control, have been well heralded. The good news is that we are not even close to running out of steam yet.

But instead of praising the microprocessor, I want to highlight some of the unintended consequences of this advantage called software.

Generally speaking, the last two generations of engineers were trained to practice the art of implementing the logic required for electronics-based systems by writing software to control microprocessors or microcontrollers. Far be it from me to even suggest this is, in itself, a problem. The problem lies in the fact that many logic implementations can be done at higher speed and lower power, when accomplished in purpose-designed hardware.

Let me offer a concrete example of exactly what I mean. The newest encryption standard from NIST is called Advanced Encryption Standard (AES), and the 128-bit version can be implemented in software driving a microprocessor or in hardware.

If we use a well-known RISC microprocessor architecture (MIPS), run it at 100 MHz and use 100 percent of the CPU, this implementation will give us almost 20 Mbits/second of encryption throughput.

Now, if we take the same RISC microprocessor and use it to manage the AES algorithm implemented in hardware, run the microprocessor at 100 MHz but only use 10 percent of the CPU, we can get almost 400 Mbits/s of AES throughput — a spectacular factor-of-20-times performance improvement. I'm not talking here about what has become a fairly common coprocessing solution, I'm talking about designing and implementing the AES engine in what we used to call discrete logic.

But what if we only need 20 Mbits/s of throughput? Then we could run the hardware solution at 5 MHz, which uses only about 5 percent of the power required for the software-only implementation. So here you see a real-world example of equal performance at significantly lower power — all based on the 1970s way of implementing logic in hardware.

By the way, if you really needed 400 Mbits/s of performance and only had the software option available, you would need to run the microprocessor at 2 GHz. Which is OK, if you like using fans in your equipment designs for cooling.

I predict we will observe a significant change in electronic equipment designs in the future, where major, well-defined tasks are offloaded from microprocessors to purpose-designed hardware, owing to both performance and power considerations. This is the way we get off the megahertz treadmill — truly a quantum change in the way electronic systems of the future will be designed. So in some ways, the future will look more like the distant past.

Tom Hart, Chairman, President and Chief Executive Officer QuickLogic Corp., Sunnyvale, Calif.