xcore®.ai Processor

XMOS' xcore.ai features 16 logical cores divided between two multi-threaded processor tiles each equipped with 512 kB of SRAM and a vector unit housing integer and floating-point ALUs, providing unmatched power, precision, and ingenuity. The ultra-low latency and highly deterministic performance make it perfect for demanding IoT applications.

Equipped with a 32-bit floating-point scalar pipeline and a 256-bit VPU, xcore.ai delivers exceptional DSP performance. With up to 1,600 MFLOPS of processing power at 800 MHz and built-in support for complex arithmetic and FFT/iFFT operations, xcore.ai empowers developers to tackle intricate DSP tasks with ease and efficiency.

The 256-bit VPU of xcore.ai supports a wide array of vector operations, enabling peak AI performance of 51.2 GMACC/s at 8-bit precision. With a unified computing environment facilitating seamless integration of AI operators, xcore.ai lays the groundwork for advanced AI/ML applications in the IoT sphere.

Combined with a unified development environment, the communication infrastructure allows for seamless integration and expansion of DSP systems by leveraging a robust interprocessor infrastructure and facilitating high-speed communication among integrated processors and multiple xcore.ai SoCs, ensuring unparalleled scalability.

xcore.ai's software-defined partitioning allows for optimal configuration of IO, control, DSP, and AI/ML functionalities, perfectly aligning with the specific requirements of each application. The ability to scale the tile clock frequency ensures precise performance matching while reducing power consumption and offering cost-effective and power-efficient IoT solutions.