The art of just enough
A valid paradigm is taking hold, delivering more compute power in the same (or lower) electrical power, cooling, and weight envelope.
Many hardware types are from the "go big or go home" school of thought (and I'd include myself in this group). It's bred into our genes. There's never enough horsepower, Al. No such thing as too many MIPS. Extra capacity, no worries – someone will figure out how to use it eventually, especially those software types down the hall (with whom I also sympathize).
But all this is about to change. It has to. We're facing the distinct possibility that the resources we've taken for granted for so long will become scarce or unaffordable in our lifetime. And our electronic designs must change to address this dilemma.
The new thinking will be the art of just enough. Not overkill, not minimalism sacrificing function for aesthetics, but just enough.
This is a course that Japanese engineering has pursued for quite some time. Not one extra penny is spent where it's not needed, although I wish someone had coughed up more than 10 cents on the lousy bracket supporting the lamp power connector in my DLP HDTV. But still, I admire the philosophy, knowing it's a challenge to optimize a design without spending any more money, time, or energy than is really necessary.
I submit that the Intel Atom processor is just enough for many embedded applications. The Atom, as we've seen in general terms, is close to a 1 GHz Celeron in processing power. The Atom chipset integrates other functions such as video streaming quite efficiently. Its performance per watt is about 8x better than the Celeron solutions it supplants.
Sure, the folks at Intel will gladly sell you more horsepower than what the Atom offers if you ask. Lots more, in fact. The question I have is: Do you really need more?
This isn't the laptop computer world, where you really don't know what software is going to be put on the beast that might break its back in terms of performance. Somebody whips out the next Crysis, and all laptops are suddenly obsolete.
This is the embedded computer world. Engineers should know what the performance expectations are because the functions are fixed. There should be few surprises; in most cases, you're in control of what applications get loaded on the machine.
Oh, I'm not suggesting imprudence. Design in sufficient margins to ensure safety, accommodate worst case, and take the necessary precautions. Don't cut corners in processing power until it could potentially hurt.
I'm certainly not opposed to using something like a Core 2 Duo or larger processor that requires more performance. I expect we'll see some great advances in multicore hardware performance and software optimization that will enable heavy-duty applications we've just imagined up until now.
But back to the question: Do you really need more horsepower than an Atom processor offers?
I can almost hear the wailing and gnashing of teeth right now. What about upgrading system performance? Someone from marketing shows up with a new application requirement demanding a lot more horsepower, but there's no money or time for a complete hardware redesign.
PC/104 and Small Form Factors readers should be able to connect the dots. If you use one of the plethora of small form factor compute modules discussed here, chances are the supplier community will have something in a compatible footprint to upgrade performance when it's needed. Pull out the old module, put in the new module, pass go, collect $200 or more, and go spend some time with the family.
And if you're concerned about finding something that meets your needs, take a look at the Atom Roundup feature on pages 18 and 19. A wide range of Atom-based products is already available, and more are on the way.
The days of ever-escalating clock speeds have, thankfully, come to an end for the most part. A valid paradigm is taking hold, delivering more compute power in the same (or lower) electrical power, cooling, and weight envelope.
The Atom processor reflects the art of just enough nicely. Let's step back and see how we can make just enough work for a change.