Xilinx reconfigurable acceleration stack provides the industry's highest computing efficiency, 2-6 times higher than other FPGA solutions
Xilinx, Inc. (NASDAQ: XLNX), a global leader in all programmable technologies and devices, recently announced at the 2016 Global Supercomputing Conference (SC 16) the launch of a new technology - the Xilinx reconfigurable acceleration stack solution, aimed at helping the world's largest cloud service providers quickly develop and deploy acceleration platforms. A FPGA based Xilinx reconfigurable acceleration stack specifically designed for cloud level applications, including libraries, framework integration, development boards, and support for OpenStack. Through the Xilinx FPGA, this reconfigurable acceleration stack solution provides the industry's highest computing efficiency: 40 times higher than x86 server CPUs; 6 times higher than competitive FPGA solutions.
Through dynamic reconfigurable technology, Xilinx achieves chip level optimization for various workloads that require high-performance computing, such as machine learning, data analysis, video transcoding, etc. By switching to the optimal design bitstream, optimization of these workloads can be completed in just a few milliseconds.
Due to the DSP architecture advantages of limited precision data types, excellent on-chip memory resources, and technological advantages over competitive FPGAs for over a year, today's Xilinx FPGAs have increased the computational efficiency of machine learning inference in ultra large scale data centers by 2-6 times.
The Xilinx reconfigurable acceleration stack includes a mathematical function library designed for cloud computing workloads, an application function library that integrates major frameworks (such as Caffe for machine learning), and PCIe for high-density servers ® The development board and reference design, as well as the OpenStack support suite, make the Xilinx FPGA accelerator easy to configure and manage.
Karl Freund, a senior analyst in the machine learning department of Moor Insights&Strategy, a market research company, said, "Flexibility and speed of functionality, that is, the ability to quickly add new features through software, are crucial for ultra large and complex application environments. As Xilinx advocates, accelerators that can be reconfigured and optimized in real time as algorithms evolve have extremely attractive advantages in ultra large scale environments
The new stack solution can accelerate the adoption of our FPGA in mainstream applications of ultra large scale data centers and provide the fastest computing efficiency 2-6 times higher than other FPGA solutions, "said Nazeem Noordeen, Vice President of the IP Solutions Division of Xilinx
2020-09-04Reading volume：6186
