AI Inference Accelerator Hailo-8™

This is a chip with an architecture specialized for AI processing.
Because it is dedicated to AI processing, the data pool and computing unit can be placed in a small area.
Therefore, it is possible to achieve processing with low latency and low power consumption without using external I/F or I/O.
As a result of the evaluation, compared to general-purpose GPUs, the area was reduced to one-fifteenth and the power consumption efficiency was 20 times higher.
In FPS comparison, performance is 1.1 times higher. (Results of our survey)

Reference architecture

The layout within the chip holds the key to improving performance.

General-purpose GPUs have separate data pools and computing units, and use external I/Fs (interfaces) and I/Os, so performance and power consumption tend to drop.

Hailo has an optimal layout and is packed into a small area, making it a highly efficient device that can be specialized for AI inference.

M.2

 

・Hailo-8™

・PCIe Interface

・M.2 2242 M-Key, B+M Key

◦2260 & 2280 mechanical compliant

・M.2 2230 A+E Key

Mini PCIe

 

・Hailo-8™

・PCIe Interface

・mPCIe 3050

○Evaluation boards can also be provided

Hailo manufacturer information Top

If you would like to return to the Hailo manufacturer information top page, please click below.