September 18, 2019 | Bård M. Pedersen
Several classes of embedded applications require large amounts of memory for executing either code or data. For example, some connected IoT nodes use wireless networking...
As Christmas approaches, thoughts naturally turn towards gifts for our friends and loved ones. While giving people animals as gifts at any time of the year is generally frowned upon, particularly by animal charities, it’s probably true that many children will still find a furry friend under their tree on Christmas day.
Taking responsibility for a pet is a major milestone in a child’s life and one that needs to be treated with the utmost importance by the child and their parents or guardians. So important is this message that in 1978 the animal charity then known as the National Canine Defence League (now known as the Dog’s Trust) coined the phrase ‘A dog is for life, not just for Christmas’. It’s a slogan that has stood the test of time, for very good reasons.
The maxim soon became deeply embedded in the psyche of people, and it continues to be paraphrased. If something is supposed to last a long time, you can guarantee someone will state ‘it’s not just for Christmas’. Having rapidly found its way into popular culture it is a reliable source for memes.
The counterargument to this truism is that some things are only expected to last a short while, however, if you ask an engineer if that applies to ASICs they may very well respond with ‘well, they’re not just for Christmas, are they!’ In the past, that may have been true, because the semiconductor industry has nurtured that interpretation. It is generally seen as an expensive undertaking to develop a full-custom IC, particularly if you are a relatively small company or have low volume demands. However, today, the reality is that ASICs don’t need to last forever to make economic sense.
Of course, there are many good reasons why a full-custom approach can be seen as a long-term commitment for engineering companies. For one thing, they offer a lower unit cost per function when measured against ASSPs or FPGAs. That’s not really surprising, because it is really why many OEMs consider a full-custom IC in the first place. However, if you suggest to these companies that they could replace their ASIC on a relatively frequent basis and still save money, they might reach for the spreadsheet in disbelief and start punching in numbers.
If you consider every ASIC in isolation and bundle all of the costs under one project then it makes assessing the pay-back period simpler; it costs this much, so we need to ship these many products for it to make financial sense. However, the pay-back on ASICs can be very short, so extending the lifetime of a device artificially and at the cost of introducing new features really doesn’t stay true to the logic behind choosing an ASIC in the first place.
In addition to this, OEMs should consider ASICs as having a product roadmap all of their own; the cost of making small changes to a full-custom design can be negligible but extend the useful lifetime of that design by many months or even years. This may require a realignment of expectations within a company’s engineering department, but should not be ignored.
When considering going down the full-custom IC route, being realistic about how long a fixed-function device can remain competitive in a constantly changing market environment is an important factor in the engineering decision process. There are many ways to extend the lifetime of an ASIC, if that is necessary, such as taking a software-defined approach where some of the functionality is managed in firmware. This would allow the functionality of the ASIC to be modified long after it has been shipped, either in the factory during production/assembly or through an over-the-air (OTA) update. In fact, because so many products are now largely defined by their firmware, an OTA approach has become fundamental to the longevity of some products. Smart connected TVs and set-top boxes, many of which run on a Linux-based software distribution, now exemplify this OTA approach to hardware configuration. The same approach can easily be implemented in an ASIC to give it an extended lease of life.
While small design tweaks or a software-defined approach can help extend the lifecycle of an ASIC, the reality is in many cases these measures may not even be required. The cost of developing and manufacturing custom ICs is on a downward trend, influenced by the economic rules that drive the semiconductor industry in general. Riding on the coattails of this trend means that taking a full-custom IC approach is likely to be much more affordable than many OEMs may think.
Unlike that puppy you received as a child one bright Christmas morning, ASICs are no longer a long-term commitment. Think of them as the toy you got, batteries included, and played with to exhaustion all over Christmas, only to relegate it to the back of the cupboard by New Year’s Day.
Happy Christmas, everyone!