CEVA goes full-HD with fully-programmable MM3000

Written by Kenton Williston February 15, 2010 Hits: 541 Favorites

0.0 (0)

CEVA’s new MM3000 is an impressive piece of IP: it provides full-HD, 1080p video decode and encode at up to 60 fps, and it is fully programmable. The performance is impressive enough by itself, but the fact that CEVA was able to pull this off in a fully-programmable platform is what really amazes me. Other HD video platforms like the POWERVR VXD and the Mali-VE6 offer only limited programmability.

The bottom line here is that CEVA has an advantage for applications that need something more than standards-based decode and encode. For example, the CEVA MM3000 supports pre/post processing algorithms such as scaling, stabilizer, and noise reduction. It can also support proprietary applications such as augmented reality, gesture recognition, and face detection.

CEVA refers to the MM3000 as it “third-generation” video engine, but the new platform bears little resemblance to its predecessors. Past CEVA video platforms have been powered by a single DSP core. In contrast, the MM3000 is a heterogeneous platform that consists of a Stream Processor and a Vector Processor. As the names imply, the Stream Processor handles bit stream coding, bit stream manipulation, and control code execution, while the Vector Processor performs filtering and vector type operations typical for pixel processing.

CEVA has been finding success with its previous-generation video engines, and I expect it to continue that successful track record with the MM3000. The platform’s flexibility and ability to support proprietary algorithms is sure to be a major selling points with SoC designers looking to add differentiating features to their products.

User reviews

There are no user reviews for this listing.

To write a review please register or login.

Written by :

Kenton Williston