The AM62A Edge AI Microprocessor is shown
in Figure 2-1. AM62A Simplified Block Diagram, is a heterogeneous processor designed for analytics
applications. There are different hardware accelerators optimized for different tasks
thus enabling an optimized power and cost footprint for different vision and multimedia
applications
The main processing and compute subsystems
from a multimedia context in AM62A are as follows:
- Quad-A53 ARM cores: These can run up
to 1.4 GHz and provide up to 16.8k Dhrystone Million Instructions Per Second (DMIPS)
of performance.
- C7 Digital Signal Processor (DSP) and
Matrix Multiplication Accelerator (MMA): TI’s deep learning accelerator on AM62A is
capable of two TOPs operations when clocked at 1 GHz.
- H.264/265 Encoder and Decoder: This
has a total encode/decode measured capability of 240MP/s and can support up to 32
channels concurrently.
- MJPEG Encoder: This has a measured
encode capability of 416MP/s and can easily support 4K30 applications.
- Vision Processing Accelerator
(VPAC3L): The latest generation in TI ISP technology for performing image operations
some examples of which are color conversions, chromatic aberration correction,
pyramid scaling and lens distortion correction. It has a total throughput of up to
300MP/s with support for RGBIR cameras.