I recently had occasion to pick up a copy of the first edition of BDTI’s Buyer’s Guide to DSP Processors, which was published 11 years ago. Flipping through the pages reminded me of how far DSP processors have come since 1994, when 3.0 volts was “low-voltage,” 50 MHz was “impressive speed,” and a 20 MHz Analog Devices ADSP-2115 sold for $21.
One thing that struck me was that eleven years ago there were actually more vendors offering general-purpose DSPs (as opposed to application-specific DSPs) in the U.S. than there are today. Of the nine vendors selling general-purpose DSPs covered in Buyer’s Guide ‘94, only three—Analog Devices, Freescale (formerly Motorola’s semiconductor business), and Texas Instruments—still do. In fact, these are the only three vendors offering general-purpose DSPs in the U.S. today. The rest of the nine from 1994 now sell only application-specific chips, or have abandoned the U.S. DSP market, or have simply disappeared.
The thinning ranks of general-purpose DSP vendors may seem surprising given that the overall DSP market today is so much bigger than it was a decade ago. One reason for the vendor reduction may be that the general-purpose DSP market is not growing as quickly as the overall DSP market. Of perhaps greater importance, however, is the trend towards highly integrated solutions. In 1994, most DSPs consisted of a processor core, some memory, and a few simple peripherals. There was nothing in the nature of these chips to tie them to one particular application area; they could readily serve a wide range of applications. For example, the same chip could be used in a cellular base station and an audio mixing console. As a result, even if you were working in a niche application you had a wide range of DSPs—and DSP vendors—from which to choose.
Today, however, chips integrate much more functionality, and much of it is application specific. There are many benefits to putting more hardware on a single chip (such as reduced system size, cost, and energy consumption). But optimizing efficiency on these metrics usually means tailoring on-chip integration to the needs of a specific application. As a consequence, most DSP vendors have narrowed their focus (and their products) to target a few applications—typically the big ones, like wireless communications and consumer multimedia. By offering chips with lots of on-chip integration, along with application-oriented software components and specialized development tools, vendors can provide products optimized for key markets, and can deliver more value. It’s like selling a complete meal rather than just an entrée.
DSP vendors’ focus on specific applications means that system developers in emerging application areas have fewer DSP processor options than they did a decade ago. However, these designers increasingly have options beyond DSPs—such as FPGAs and DSP-enhanced general-purpose processors. Some of these alternative technologies offer very strong signal-processing performance, and some (like FPGAs) are more flexible than DSPs, which can be a significant advantage in application areas with rapidly changing requirements.
A handful of today’s niche signal processing applications will become tomorrow’s killer apps. An application that starts out using an alternative technology is unlikely to later migrate to a DSP, even if volumes become large. Thus, by abandoning the general-purpose DSP market, DSP processor vendors risk ceding future high-volume applications to alternative technologies.
Add new comment