CEVA Boosts TeakLite Speed

Submitted by BDTI on Mon, 01/24/2005 - 21:00

This month CEVA introduced the CEVA-TeakLite-II, a derivative of its widely licensed CEVA-TeakLite DSP core. The CEVA-TeakLite-II differs from its predecessor mainly in terms of speed. According to CEVA, a CEVA-TeakLite-II achieves a worst-case clock speed of 200 MHz, making it roughly 30% faster than the CEVA-TeakLite. (All performance and area figures in this article are for the TSMC CL013G process.) Interestingly, CEVA was able to achieve this increase in clock rate without increasing the length of the pipeline. Instead, CEVA says it sped up the design by tweaking certain elements of the design. For example, the new core initiates modulo addressing operations earlier in the pipeline. This allows addressing to be spread over multiple cycles, enabling a higher clock speed.

CEVA also expanded the addressing range and added debug hardware in the new core. The CEVA-TeakLite-II can address 4 Mbytes of memory, compared to 256 Kbytes for the original CEVA-TeakLite. And while the CEVA-TeakLite was provided without any debug hardware, the CEVA-TeakLite-II includes debug hardware and a JTAG module.

Like its predecessor, the CEVA-TeakLite-II primarily targets cellular handsets, hard disk drives, and portable audio players. With its speed boost and larger memory space, the CEVA-TeakLite-II also adds low-end imaging and video processing to its list of target applications.

The CEVA-TeakLite-II will compete with other low-end DSP cores like the ZSP200. BDTI’s preliminary analysis shows that the CEVA-TeakLite-II is likely slower than these competitors. However, the CEVA-TeakLite-II is also one of the smallest DSP cores available. According to CEVA, the core occupies a die area of about 0.4 mm2. In comparison, other DSP cores typically occupy 1 to 3 mm2. Hence, the CEVA-TeakLite-II may be attractive for highly cost-constrained applications that need only modest processing speed.

The CEVA-TeakLite-II will also compete with low-end general-purpose processor (GPP) cores like the ARM9. For signal processing tasks, the CEVA-TeakLite-II has some notable advantages over these GPPs. It is significantly faster than competing GPPs and (based on data provided by CEVA) is also much more energy-efficient than competing GPPs. In addition, the CEVA-TeakLite-II die area is about the same as that of a low-end GPP.

The CEVA-TeakLite-II is binary backwards-compatible with CEVA-TeakLite as well as the older CEVA-Oak core. The Oak/TeakLite family has been remarkably long-lived and successful. CEVA-Oak was introduced a decade ago, and since that time 50 companies have licensed CEVA-Oak or CEVA-TeakLite. With the addition of TeakLite II, CEVA is likely to extend this impressive track record.

The CEVA-TeakLite-II is currently available for license; contact CEVA for information on licensing fees and royalties.

Add new comment

Log in to post comments