At September’s Intel Developer Forum, Intel President and COO Paul Otellini gave a keynote speech that lacked the usual obsession with clock speed. Instead, Mr. Otellini’s speech focused on issues like parallelism, integration, and power consumption. What struck me about this change in emphasis is that Intel now seems to be reading from an embedded processor vendor’s playbook.
Take Intel’s perspective on parallelism. Mr. Otellini noted that Intel plans to move all of its x86 families to multi-core architectures. One motive for this dramatic shift is the so-called “thermal wall.” Intel has acknowledged that the clock speeds of its CPUs are now limited by power constraints. The only way to make chips faster without unacceptable increases in power is to boost parallelism. Conversely, parallelism provides a means to lower power consumption in products that don't need to run at top speed. Perhaps that explains why Intel expects 70% of new laptops—compared to just 40% of desktops—to contain dual-core processors by the end of 2006.
While parallel processing may be big news for the PC community, embedded applications—particularly signal-processing applications—have relied on parallel processing techniques for decades. And companies like Freescale and Qualcomm already offer DSP-oriented processors with four or more cores, compared to only two cores in Intel’s latest PC processors.
Intel’s vision for future PC software should also sound familiar to embedded system developers. Mr. Otellini suggested that one processor core could be used for the user interface, while the other could handle number-crunching tasks. That’s the same basic approach used in embedded applications ranging from cell phones to set-top boxes.
What interests me most about Intel’s new worldview is that it indicates a major shift in the way processors are developed. Until recently, leading-edge technologies were developed for PCs and then slowly migrated to embedded applications. But as the demands and importance of embedded applications have grown, embedded applications have taken the lead in some areas. As a result, advances in processor technology now seem to be flowing in both directions.
This increasing cross-fertilization between PC and embedded technologies creates exciting possibilities. For example, Intel’s dual-core chips need an OS that can use both cores. What if the same OS could be used on multi-core embedded processors? It’s too soon to say whether such synergies will arise, but it’s certain that embedded applications are no longer waiting for technology to trickle down from the PC world.
Add new comment