Media consumption and SoC design: envisioning the media SoC of tomorrow

A panel at the 2008 Globalpress Summit in San Francisco this week examined "Consumer Content and Media," and ended up with some interesting implications not just for how consumers will absorb media, but how those patterns will influence IP requirements and SoC design in the consumer electronics world.

The vision suggested by panel moderator Tim Bajarin, principal analyst at Creative Strategies, is of an internet-based cloud of content, floating free of any particular content source, gateway, or repository. Bajarin suggested that as consumers became accustomed to this nebulous new landscape, they would freely view content in the cloud from any of the three gradually-merging windows: the PC screen, the TV screen, and the handset screen.

Conspicuously absent from this vision is the strong sense of hierarchy that has been the foundation of a lot of thinking on home media gateways and networks. The ultimate implication of Bajarin’s view, supported to some degree by his panelists, is that there is no hierarchy of networks. If there is a local network within the home—and as a matter of fact, Bajarin suggested that UWB networks might well fill this role—it is not so much an isolated pathway between a private media server and the home’s three kinds of screens as it is a local, high-bandwidth extension of the Internet itself: merely another wisp of the cloud.

Nor did the panel assume a hierarchy of content creation. "This is a new way of thinking about content," maintained Stuart McKechnie, director of strategic marketing at camera chip maker Zoran. He suggested that the new landscape is not a road from Hollywood downhill to the TV screen. "Imagine your camera and your TV directly connected to the cloud. You don’t have access to 200 channels, but 200,000. Or maybe 3 million—the consumers become creators of content." McKechnie mentioned in passing that Zoran is already supplying silicon to at least a couple of digital cameras capable of being direct Web clients.

The panel also discussed Hollywood’s struggle to understand and find a business model in this new stratosphere. There was general agreement that as a matter of survival, silicon vendors would have to support whatever content-protection mechanisms Hollywood demanded, whether they made business sense–or even whether they worked–or not.

Taken together, the panel’s ideas present a very interesting set of implications to the CE SoC design team. In the first place, it appears that in this view of the world, all devices capable of capturing or displaying content will be directly connected to the Internet. But that connection will have to be agile: any of a number of UWB approaches (represented on the panel by Pulse-LINK CEO Bruce Watkins) or 802.11 within the home, and, by Bajarin’s list, LTE, true 4G networks, and WiMAX. Time, content, and circumstance determine the choice of network, not a priori design.

Agility also applies to a number of other key—and compute-intensive—tasks. Content from the cloud may come compressed in a format appropriate for the material, for the intended display, or simply for the convenience of the creator. There will be no one—or two, or three—standard media compression formats, but potentially dozens, each of which each screen must handle. No one on the panel expressed hope that a single scalable video CoDec would be the one solution to all content-to-display match-ups.

Similarly, there will likely be not just a set, but an ongoing countermeasures war of content protection and authentication schemes. Observing that both DVD and BlueRay content protections were cracked almost at once, McKechnie predicted a continuing war, forcing display devices to adapt to a monotonically growing compute load just to protect other people’s presumed intellectual property.

This scenario suggests a picture very different from the usual maturity curve for silicon products, which traditionally begin highly configurable and over time become more compact, efficient, and fixed-function. Rather, we see not the gradual hardening and shrinking of media SoCs—whether set-top, PC-based, or handset-resident—but rather their continual evolution, always having to adapt to one more new compression scheme, network interface, or security mechanism.

The panelists had at least two views on how this could be possible. One, expressed by Steve Roddy, vice president of marketing at Tensilica, predictably favored an array of compact, optimizable CPU cores to give both adequate flexibility and adequate performance at acceptable power levels. The other, implied but not argued by McKechnie, was also a multicore approach, but one that relied much more heavily on hard-coded blocks to reduce energy consumption.

Either approach, whether you are optimizing CPU cores with extended instructions, registers, and I/O paths; or whether you are creating dedicated hardware engines; requires an artful job of identifying the high-compute-load kernels across a wide range of algorithms, isolating these kernels, and parallelizing them. This in itself will become a primary challenge in SoC design, especially given that algorithm developers will attempt to differentiate themselves specifically by modifying the kernels. But the SoCs must anticipate, not follow, the algorithms, if the vision of the ubiquitous media cloud is to work as a business model. To the art of SoC architecture, add the sideline of prognostication.