In recent years, ARM has come to dominate the market for licensable general-purpose processor (GPP) cores. Its main competitor, MIPS, has fallen into a distant second place and most other GPP core licensors are niche players. ARM has become so dominant that its competitors are unlikely to threaten its number-one position in the near future.
At first glance, a similar trend seems to be developing in the market for digital signal processor (DSP) cores. Over the last few years, the competitive field has consolidated from about a dozen licensors to just three major players. Will one of these vendors eventually gain an ARM-like unassailable lead?
I wouldn't bet on it. The market for DSP cores is remarkably different from the market for GPP cores. For one thing, compatibility is less important in signal-processing applications than in other applications. This is reflected in the different approaches taken by ARM and by CEVA, the largest DSP core vendor. ARM's architectures have maintained binary compatibility for over a decade, while CEVA's latest offering is incompatible with its predecessors. The lower importance of compatibility makes the DSP core market more open to competition. And CEVA's two primary competitors, LSI Logic and StarCore, have significant resources and the potential to become market leaders.
In addition, ARM itself may prevent any DSP core vendor from becoming too strong. As GPP cores become faster, they are able to handle an ever-broader range of signal-processing tasks. In contrast, it is difficult for DSPs to take on some GPP tasks such as running an OS. Many DSPs have all the features needed to run an OS, but they face a software barrier: OS support for DSP cores is fairly limited. As a result, GPPs are stealing some signal-processing applications away from DSPs.
The growing use of GPPs for signal-processing tasks raises another interesting question: will ARM eventually displace DSP core vendors as well as its GPP competitors? Again, I think the answer is no. Although GPPs are capable of handling many signal-processing tasks, they are not necessarily efficient at these tasks. For applications where cost-efficiency or energy-efficiency is key, DSP cores often hold a distinct advantage.
In short, competition is alive and well in the DSP core market. The market cannot support a dozen successful suppliers, but I think there will be healthy competition for years to come.
Add new comment