+1 607 252-6647 Email Questions

​Samsung boosts mid-tier Exynos 7 AP with deep learning image processing

  • Published in News
  • Hits: 411

Samsung Electronics has launched a mid-tier mobile application processor (AP) with boosted deep learning and image processing capabilities.

The Eyxnos 7 9610, aimed at mid-tier smartphones, is made through the 10-nanometer process and has strengthened image processing capabilities that meet the phone trend of high camera and multimedia usage, the South Korean tech giant said.

Thanks to deep learning technology and optimised hardware, the new AP can recognise faces when taking a picture even when the subject's face is obstructed by hair or hats, or if they are not facing the camera.

Depth detection has been made more precise to allow better out-focusing, even for single cameras, Samsung said.

Eyxnos 7 9610 supports codec that can film 4K 120 frames per second (fps), and has 1.6 times faster image singular processing speeds compared to its predecessor, Exynos 7 7885. These capabilities first featured on the Eyxnos 9 9810[1], Samsung's premium AP, which the company also uses for its own smartphone models.

Its mobile industry processor interface, which connects the AP with the image sensor, is two times faster. This allows the filming of 480fps Full HD slow motion video even if a 2-layer image sensor is used.

The octa-core processor has four 2.3GHz Cortex-A73 big cores and 1.6GHz Cortex-A53 little cores, and an Bifrost-based ARM Mali-G72 GPU. It has a 6 mode modem that covers 2G CDMA to 4G LTE, and has download speeds up to 600Mbps and upload of 150Mbps.

Samsung said it is expected to be mass produced in the second half of this year. Mid-tier products like Exynos 7 are largely aimed at the Chinese market.

Samsung's logic chip business is diversifying its portfolio. Last year, the firm began

Read more from our friends at ZDNet

Contact us

By Mail

PO Box 5613

Katy, TX 77491

USA

Social: twitter facebook

Phone: +1 607 252-6647

Fax: +1 866 573-1096

Email: info [AT] synapticweb [DOT] co