Modern, high performance Audio DSP, optimized for far-field noise reduction and Artificial Intelligence speech recognition

Overview

The Ceva-BX2 audio/voice DSP is targeted for high performance audio devices such as DTV, Smart Speaker, Soundbar, and car infotainment systems.
Ceva-BX2 uses quad 32X32-bit MACs and octal 16X16-bit MACs, with enhanced capability for supporting 16×8-bit and 8×8-bit MAC operations.

The Ceva-BX2 is using an 11-stage pipeline and 5-way VLIW micro-architecture, it offers parallel processing with dual scalar compute engines, load/store and program control that reaches a speed of 2 GHz at a TSMC 7nm process node.

The Ceva-BX2 Instruction Set Architecture (ISA) incorporates support for Single Instruction Multiple Data (SIMD) as well as optional floating point units for high accuracy algorithms.
The Ceva-BX2 is accompanied by a comprehensive software development tool chain, including:

* Advanced LLVM compiler
* Eclipse based debugger
* DSP and neural network compute libraries
* Neural network frameworks support
* Real Time Operating Systems (RTOS)

Key Features

Octal 16x16 MACs
Quad 32x32 MACs
5-way VLIW
8/16/32/64-bit data types
16x8 and 8x8 Neural Network support
Half and single precision IEEE floating point units
Innovative Branch Target Buffer minimizing branch overhead
Hardware loop buffer for reduced power consumption of code loops
High performance controller
- 5 CoreMark/MHz
- Dynamic branch prediction
- Full RTOS support
- Compact code size
Advance system control
- Automatic Queue and Buffer management mechanisms to integrate co-processors and create a cluster of Ceva-BX cores
- Dedicated HW accelerator ports

Benefits

Up to 16 GMAC/sec tailored for sound neural networks
High-throughput DSP addresses advanced applications
Ceva-Connect offloads the processor from data transfers to hardware accelerators and peripherals

Block Diagram

Technical Specifications

Maturity

In Production

Availability

Available

TSMC

Pre-Silicon: 12nm

Request Info