MYTHIC - 模拟矩阵处理器加速边缘人工智能

AI的巨大潜力并没有发挥出来 -

• 当今的数字AI处理器成本高昂，且使用起来非常复杂(overly complex)
• 采用数字处理器做推理(inference)面临着巨大的技术挑战，如尺寸，性能和功耗等

MYTHIC为边缘侧设备带来先进的人工智能推理。

• 其AMP(Analog Matrix Processors，模拟混合处理器，带来巨大的功耗，性能和成本优势。
• AMP在高密度闪存阵列(flash-memory array)采用模拟计算(analog computing)执行深度神经网络推理计算。
• 提供高级程度开发环境(integrated development environment)，可快速部署复杂的深度神经网络(the most sophisticated deep neural networks)

广泛的应用场景 -

MYTHIC开发的独特的AI技术平台，用AI赋能包括surveillance camera systems, smart appliances, machine vision systems, commercial drones, robotics等多种产品。拥有独一无二的性能和功耗优势 -

• Lowest latency: Single frame delay by running batch size of 1
• Efficient Compute: Best-in-class performance-to-power efficiency
• Hyper scalability: Low-power single-chip to high-performance rack systems
• Ease of use: Supports major platforms, frameworks and topology-agnostic performance

MYTHIC产品采用三项基本的硬件技术，分别是 -

Compute-in-memory

Dataflow Architecture

Analog Computing

不同于传统的计算架构 -

AI推理不同于典型的串行应用(typical sequential application)，更类似于图像处理的并行计算，主要是加法和乘法(multiplication and accumulation)。

模拟计算架构 -

This is possible by using the memory elements as tunable resistors, supplying the inputs as voltages and collecting the outputs as currents. We use analog computing for our core neural network matrix operations, where we are multiplying an input vector by a weight matrix.

MYTHIC的AI推理工作流程 -

更多信息请参考如下视频 -