Stronger power for AI

Parallel computing chips with high versatility and high performance,  become a leader
in the filed of parallel computing

RPP reconfigurable chip

Chips based on RPP architecture will better meet the high-performance
computing needs of various scenarios

Professional R&D team

The core team has more than fifteen years of professional R&D experience in the
field of  chip design and development

Stronger power for AI

Parallel computing chips with high versatility and high performance,  become a leader
in the filed of parallel computing

RPP reconfigurable chip

Chips based on RPP architecture will better meet the high-performance
computing needs of various scenarios

Professional R&D team

The core team has more than fifteen years of professional R&D experience in the
field of  chip design and development

Stronger power for AI

Parallel computing chips with high versatility and high
performance, become a leader in the filed
of parallel computing

RPP reconfigurable chip

Chips based on RPP architecture will better meet the
high-performance computing needs of various scenarios

Professional R&D team

The core team has more than fifteen years of
professional R&D experience in the
field of  chip design and development

Our team

Innovative chip architecture

Our vision

ABOUT US

AzurEngine has developed an innovative processor architecture called RPP (Reconfigurable Parallel Processor), which bridges the gap of programmability and power efficiency between GP-GPU and ASIC (TPU/NPU).

RPP is a high-performance, efficient, and compatible chip solution for parallel computing at the edge and in the data center.

Small and Powerful

Our innovative RPP architecture enables high-efficiency, general-purpose, and low-power GP-GPU chips.

Existing GPU solutions are pushing the limits of Si area and system power consumption, mostly relying on process technology advancements for PPA improvement.

RPP-based chips provide a fully programmable solution with 3x reduction in Si area, improved performance, and lower power consumption.

AzurEngine RPP architecture

1024 RPP computing unit

 

NVIDIA Volta architecture

1024 GPU computing unit

Chip area
< mm²
Resnet50 performance
> ips
Chip power consumption
< W

Balance performance and versatility

Compared with the vector processor architecture, the RPP architecture will provide extremely high computational efficiency. Compared with the current mainstream GP-GPU,
The RPP innovative architecture will bring higher computing density and versatility close to GP-GPU, as well as efficiency comparable to professional chips

Core Products

In recent years, with the rapid development of AI technology, the demand for deep learning-based computer vision applications such as face recognition, car recognition, and logistics inspection has increased rapidly. Faced with the current and future requirements for chip computing power and programming flexibility With rapid growth, the market urgently needs powerful, flexible and low-power parallel computing chips.

GP-GPU chips based on the RPP (Reconfigurable Parallel Processor) architecture can effectively reduce chip costs while taking into account versatility and computing performance, and better meet the market’s long-term demand for chip products.

Application scenario