Xilinx Invests in DeePhi Tech, Machine Learning Specialist
Published2017.07.06 10:43Xilinx announced its investment in DeePhi Tech, a leading player in the machine learning field specializing in deep compression, compiling toolchain, and system-level optimization.
DeePhi provides an inference platform for next-generation AI products and services across the industry from edge to cloud by leveraging the structural advantages of Xilinx® devices in machine learning.
DeePhi Tech was founded by researchers from Tsinghua University and Stanford University, leading institutions in machine learning research, and has published numerous papers at the world's top AI conferences. Recently, DeePhi Tech members received the best paper award at FPGA 2017 for their research on a high-efficiency FPGA-accelerated speech recognition engine that achieved 43x performance per watt and 40x performance compared to CPU, and 3x performance and 11x performance compared to GPU.
Yao Song, CEO of DeePhi Tech, stated, "Xilinx's FPGA and MPSoC provide an ideal combination of short latency and reconfigurability required for our inference solutions," and added, "Xilinx's powerful platform is already being used in consumer products ranging from edge to cloud in smart surveillance systems and data center applications."
Nazeem Noordeen, CVP of Xilinx Acceleration and IP Solutions, said, "DeePhi's solutions, combining Xilinx's technology and expertise, improve efficiency based on technologies such as network pruning, compression, and the use of reduced-precision data types," and noted, "Xilinx's investment in DeePhi Tech represents Xilinx's commitment to providing customers with more industry-leading accelerated computing products."
Related news
Xilinx and IBM Improve Connection Performance Between Accelerators and CPUs with PCI Express Gen4
Xilinx announced achievements in PCI Express® Gen4 functionality. Xilinx, in collaboration with IBM, has doubled the interconnect performance between accelerators and CPUs through PCI Express Gen4 compared to the previously widely-used PCI Express Gen3 standard. Gen4 accelerates performance in data center applications such as artificial intelligence and data analytics by increasing bandwidth between CPUs and accelerators to 16Gbps per lane. IBM and Xilinx implemented Gen4 interoperability between Xilinx® 16nm UltraScale+™ devices and IBM POWER9 processors, demonstrating PCIe Gen4 functionality for the first time in a programmable device.
2017.06.29by 김지혜 기자
Xilinx Integrates Flexible Transceiver Technology for Optical and High-Performance Interconnect
Xilinx announced the integration of 56G PAM4 transceiver technology into the Virtex® UltraScale+™ FPGA. Built on and verified in 16-nanometer FinFET+ FPGA fabric, this device extends the Virtex product line to deliver next-generation Ethernet deployment and enables flexible migration of existing systems to next-generation backplane, optical, and high-performance interconnect. Using integrated devices targeting wired communication, data center, and wireless backhaul applications, customers can overcome the physical limitations of data transmission at 56G+ line speeds, thereby doubling bandwidth in existing infrastructure.
2017.07.04by 김지혜 기자
ST Microelectronics and NUS Establish Joint Research Center 'HELIX' for Edge AI
ST Microelectronics and the National University of Singapore have established the 'HELIX' joint research center aimed at implementing embedded AI in edge environments. Over a four-year period, the center will conduct research across the entire technology stack from algorithms to silicon, focusing on memory-centric architecture leveraging ST's 18nm FD-SOI technology and embedded PCM. The core objective is to implement intelligence directly in physical devices such as robots and drones while minimizing energy consumption.
2026.08.28by 배종인 기자
ams, Enhancing Humanoid Robot Sensing and Communication with Digital Photonics
ams OSRAM has unveiled solutions that leverage digital photonics technology to enhance the sensing and communication capabilities of humanoid robots. The company implements environmental awareness and precision manipulation through dToF sensors, optical force sensors, and structured light emitters, while supporting LED-based non-verbal communication through the EVIYOS and ALIYOS platforms. According to market research, the humanoid robot market is projected to grow from $600 million in 2025 to $51 billion in 2035, with improvements in robot sensing capabilities expected to serve as a key growth driver.
2026.08.27by 배종인 기자
Nvidia Unveils Jetson Orin Nano 2…78TOPS·2x Inference Performance
Nvidia has unveiled the Jetson Orin Nano 2. It delivers 78TOPS AI computing performance and 2x inference performance compared to the previous model while reducing power consumption by 40%. The device is scheduled for launch in the first half of 2027, and major robotics and drone manufacturers including Cognex, Doosan Bobcat, Matic, and Wing are currently reviewing early adoption.
2026.08.26by 배종인 기자
