OVERVIEW
With its unique multi-threaded micro-architecture, xcore.ai provides low latency with highly deterministic performance, ideal for intelligent IoT applications. Each xcore.ai features 16 logical cores split between 2 multi-threaded processor ‘tiles’. Each tile is equipped with 512kB of SRAM and a vector unit providing integer and floating point ALUs. A powerful interprocessor communication infrastructure provides high speed communication between each integrated processor, and multiple xcore.ai SoCs, for ultimate scalability. All this, within a single, homogeneous, and powerful development environment.
WHITEPAPER AND REPORTS
APPLICATIONS
Architected to deliver high processing and low eBOM. Easy to use, cost effective and scalable. It’s designed for smart, connected products across multiple applications.
FEATURE HIGHLIGHTS
Software-Defined IO
- Deterministic processing performance
- Precise ns timing accuracy
- Providing fully programmable application specific interfaces
- Embedded PHYs for MIPI, USB and LPDDR
Control Processing
- SMP FreeRTOS across multiple threads
- C/C++ code development environment for control code integration
- Up to 3200 MIPS of performance available on the 800MHz package options
Flexibility
Software-defined partitioning of xcore.ai threads provide optimal configuration of IO, control, DSP and AI/ML that perfectly matches the specific requirements of each target application. Scaling of the tile clock frequency to meet exact performance requirements reduces power consumption providing cost and power effective intelligent IoT solutions.
DSP Performance
- 32-bit floating-point scalar pipeline offering up to 1600 MFLOPS of performance at 800MHz
- 256-bit VPU adds block floating point capabilities up to 12.8 GMACS of performance at 800MHz
- Integrated complex arithmetic and FFT/iFFT support at up to 1 million 256-point FFT/s
- view more about xcore.ai DSP capabilities
AI / ML Performance
- 256-bit VPU supports 32-bit, 16-bit, 8-bit and binary vector operations
- Peak AI performance of 51.2GMACC/s at 8-bit with a sustained performance of 40.96GMACC/s at 800MHz
- Homogeneous computing environment for tight integration of AI operators
LET’S GET STARTED
Getting started on xcore.ai is easy! Try one of our ready-made solutions, or build your own.
VOICE PROCESSING
xcore.ai’s unique architecture combined with the XCORE-VOICE framework allows designers to incorporate keyword, event detection, or advanced local dictionary support to create a complete voice interface solution.
AUDIO PROCESSING
The xcore.ai multichannel audio platform provides a scalable and flexible hardware and software solution for a wide range of consumer and professional audio products.
VISION PROCESSING
Our solution combines the fast, flexible and economical xcore®.ai, and a light neural network model with licence plate detection and optical character recognition
BUILD YOUR OWN SOLUTION
Our xcore.ai evaluation kit allows testing in multiple application scenarios and provides a good general software development board for simple tests and demos. Click below to get started.
BUY
| Part Number | Package | IO voltage | IO | External Interfaces | Datasheet | Buy |
|---|---|---|---|---|---|---|
| XU316-1024-QF60A | 60pin QFN (7x7mm)​ | 1v8​ | 34 | USB | DATASHEET | |
| XU316-1024-QF60B | 60pin QFN (7x7mm)​ | 3v3​ | 34 | USB | DATASHEET | |
| XU316-1024-FB265 | 265pin FBGA (14x14mm)​ | 1V8 / 3V3​ | 128 | USB, Single or Dual lane MIPI D-PHY receiver​, LPDDR1 | DATASHEET | |
| XU316-1024-TQ128 | 128pin TQFP (14x14mm) | 1V8 / 3V3 | 78 | USB, Single or Dual lane MIPI D-PHY receiver | DATASHEET | |
