These days, there are so many start-ups developing programmable processors that it feels like we’re back in the “bubble” years, when anyone with a remotely viable processor design could secure venture funding. A pivotal question for the current crop of start-ups is whether to offer their processors as flexible, general-purpose chips, or as highly specialized, application-specific solutions. Should it be a jack-of-all-trades, or a master of one?
If the processor is complex or the programming tools are iffy, vendors may be more successful if they offer a full application-specific solution—complete with robust application software and hardware reference designs. This way, their customers can enjoy the power of thenew device without having to learn all the nitty-gritty details of the architecture and tools. But this approach requires deep application expertise and system design skills—and many processor companies don't have these kinds of resources. A narrow market focus can also be risky: there are fewer potential customers, and fewer shots at success.
If the processor is easy to use, the vendor may be better off going the general-purpose route, by offering strong support for application developers and a range of chips with different on-chip integration. This way the company can target a wide range of customers and perhaps increase the odds of getting its chip embedded in the next killer app. Of course, this strategy can be expensive; it requires solid development tools and the ability to support a diverse range of customers.
Depending on the vendor, either strategy can make sense. Unfortunately, we often see processor start-ups that haven't given sufficient thought to which approach they will take. A common trajectory goes like this: A vendor introduces a mind-numbingly complicated chip, claims that it's ideal for every application under the sun, and waits for customers to start queuing up. But customers are terrified of the unusual processor and won't take a chance on it. The vendor belatedly realizes that, to get anyone to use its chip, it’s either going to have to provide a more complete, application-specific solution, or make the chip easier to use. Both of these approaches take time and money, and too often, the vendor runs out of one or the other before fielding a really viable product.
Add new comment