Xilinx, Inc. (NASDAQ: XLNX), a global leader in adaptive and intelligent computing, recently announced the launch of a breakthrough new product, called ACAP (Adaptive Compute Acceleration Platform), which surpasses the functions of FPGA. ACAP is a highly integrated multi-core heterogeneous computing platform, which can be flexibly modified from the hardware layer according to the requirements of various applications and workloads. ACAP has the self-adaptive ability of dynamic adjustment in the working process, which realizes the performance and performance power consumption ratio that CPU and GPU cannot match. Now let kywos introduce the details of ACAP products:

1.ACAP product background

In the era of rapid rise of big data and artificial intelligence, ACAP is ideal for accelerating a wide range of applications, including video transcoding, database, data compression, search, AI inference, genomics, machine vision, computing storage and network acceleration. Software and hardware developers will be able to design ACAP based products for endpoint, edge and cloud applications. The first ACAP product series will be "Everest" product series developed by TSMC 7 nanometer technology, which will be streamed at the end of this year.

2.ACAP Technical Details

The core of ACAP is a new generation of FPGA architecture, which combines distributed memory with hardware programmable DSP modules, a multi-core SoC, and one or more software programmable computing engines with hardware adaptability, all of which are interconnected through the network on chip (NoC). ACAP also has highly integrated programmable I/O functions, ranging from integrated hardware programmable memory controller to advanced SerDes transceiver technology, cutting-edge RF-ADC/DAC and integrated high bandwidth memory (HBM) according to different device models.Software developers will be able to use C/C++, OpenCL, Python and other software tools to apply the ACAP system. At the same time, ACAP can still use FPGA tools to program from RTL level.

3.Performance improvement of "Everest"

With the latest 16nm Virtex ®  Compared with VU9P FPGA, "Everest" is expected to improve the performance of deep neural network by 20 times! The 5G remote radio head end based on "Everest" can increase the bandwidth by 4 times compared with the latest 16nm radio. At that time, various applications across multiple markets, including automotive, industrial, scientific and medical, aerospace, testing, measurement and simulation, audio, video and broadcasting, and consumer electronics markets, will achieve significant improvements in performance and power efficiency.

          To make a long story short, Victor Peng, president and chief executive officer (CEO) of Xilinx, said: "This is not only a major technical disruption to the industry, but also our most outstanding engineering achievement since the invention of FPGA. This revolutionary new architecture is part of the broader market strategy of Xilinx, which will help the company develop beyond FPGA and break through the limitation of" only supporting hardware developers ". The application of ACAP products in the data center and our broad market will accelerate adaptive computing technology ", so that the intelligent, interconnected and adaptive world can become a reality earlier."