In the last year or so I have noticed an impressive surge in new processors targeting digital video applications. Hardly a month goes by without a vendor announcing that it has the ultimate video processor. Countless new companies have sprung up to address this hot market, and established processor vendors are scrambling to re-spin their offerings for digital video.
Why is everyone suddenly so eager to market a digital video processor? One reason is that digital video products show great promise. Applications ranging from home entertainment to surveillance are quickly going digital—and are growing quickly—creating huge markets for digital video chips. And we’re just getting to the point where programmable processors have enough performance to run demanding video applications—so processor vendors are getting their first opportunities to access these markets.
Another reason for the rise of video processors is that there are many video compression/decompression (codec) standards, and no one standard has taken a commanding lead. When it’s not clear which video codec your product should use—or if it must support multiple codecs—the flexibility of a programmable processor becomes very attractive.
But what if video codecs some day stabilize? If everybody settles on the same codec, why bother with a programmable processor? For one thing, video compression standards don’t specify all of the inner workings of the video compression algorithm. Clever engineers can—and do—come up with ways to tweak the encoding algorithm to improve the compression ratio or to produce better-looking output. A programmable processor allows such tweaks to be implemented without the cost, risk, and time required for custom silicon.
Perhaps more importantly, video processing involves more than compressing and decompressing data. While many video systems do use compression, they also typically include pre- and post-processing steps such as de-interlacing, scaling, and color correction. Different companies employ different algorithms for these pre- and post-processing steps, giving system designers an opportunity to differentiate their products. And it’s easiest to implement unique pre- and post-processing algorithms on a programmable processor.
For these and other reasons, it is clear that there will be a large market for video processors for many years. What’s not so clear is whether this market has enough room for the seemingly endless stream of new processors. In many mature applications, only two or three architectures have meaningful market share—will the same thing happen in digital video?
Perhaps not. For one thing, digital video applications are remarkably diverse. To cite a few examples, HDTV applications demand blazing processing speed, mobile video applications demand excellent energy efficiency, and surveillance applications demand extremely high reliability. This suggests that there is room for lots of digital video architectures, because there are lots of different problems to solve.
On the other hand, fielding a successful new processor is getting harder in many respects. Digital video products are complicated systems, and they are growing more complicated over time. As a result, it is no longer enough for a vendor to deliver a processor and some basic software tools. Rather, processor vendors are increasingly expected to provide extensive application-specific development support, including things like optimized software components and reference designs. Building up this infrastructure requires a huge investment. Many companies can’t—or won’t—make the needed investment.
So while digital video markets present opportunities for many processors, succeeding in these markets isn’t easy. To succeed, vendors must focus on a particular market segment, and must deliver the right architecture, tools, and software components for that market. Without this kind of focused, holistic approach, vendors risk getting lost in the shuffle.
Add new comment