New performance and graphics demands by mobile users are forcing architectural changes in SoCs; low power plus better performance now required.
By Pallab Chatterjee
Mobile devices today are optimized for low- to mid-bandwidth data transmission, which is sufficient for e-mail, batch downloads of applications and music, and playback of encoded/compressed streaming video. But in coming months they will add another feature—image capture and processing and advanced graphics processing.
This adds a whole new wrinkle to mobile devices, which have focused more on simply downloading and displaying data—particularly at low power—than on the quality of the display and user experience—both of which require significantly higher performance.
The semiconductor industry from its inception was driven by increasing performance and increasing device density. In the past decade, the realization that power was not unlimited and needed to be part of the design constraint, has changed the direction for processors and memory structures. This swing to ultra low power is now swinging back to a combination of performance and low power.
Mobile devices are now falling into the following bins: single-processor smartphones for email/Web surfing/music/simple apps, and dual-core processors for the new high-res business class smart phones and the current line of tablets. These dual-core systems also are being used to run new gaming software with higher-resolution graphics, more movement and action in the games (not just side scrolling) and more advanced sound or motion control. This advanced software, just like on the PCs and console platforms, is required to meet user demand for even more advanced experiences, but that requires even more processing power.
That explains why the next-generation tablet processors are quad-core systems. Nvidia is currently ramping a quad-core Tegra chip, which is both performance and power optimized. The quad core allows not only for multitasking with the Android operating system, but also for high performance graphics for single tasks. The new processor is power optimized to provide the same net battery operation life as the dual core, while displaying both streaming video and high-resolution gaming.
Mobile devices have added image capture (cameras) in the past. The new generation has both outward and user (inward) facing cameras, and some have three cameras so the outward facing supports Stereo3D images. To address this level of image capture, both still and video, suppliers like Samsung are providing local image processing control circuits. These circuits handle the image stabilization, focus, zoom, white balance, among other image cleanup functions.
A new function of these mobile systems is to not only capture the image, but also edit the data. This involves intense graphics processing to provide color changing, cropping and overlay for still images. For video it requires rudimentary title insertion, NLE functions, cut and paste, and most important, encoding the edited file to a playable format.
For the gaming environment, the mobile platforms have to write code specifically for the video cores. This direct control is to be able to minimize latency and control the power management options of the graphics chip. The current methods for 2D and dimensional 3D games involve creating Open GL and Stereoscopic 3D with Open CL.
Finally, for storage in the mobile devices, the method of choice is solid state disk (SSD) technology. These devices dominate with the SATA II interface at this time. The high speed SATAIII (6Gb/s interface) is not really used in mobile devices due to power consumption issues. The SATA format gives the developer the ability to use a known timing model for the disk interface, and the SSD provides a minimum systematic latency for the disk access. The SSD has far superior power over a traditional rotating media HDD and is fairly independent on performance from single through quad core processors.
Leave a Reply