Chinese with a bold challenge to the huge Nvidia advantage in artificial intelligence (AI) hardware Researcher Trained cutting-edge video generation models with industrial chips on the shelf. It outperforms high-end GPUs in both speed and efficiency.
Their system, FlightVGM, recorded a 30% performance improvement and had more than 4.5 times more energy efficiency. nvidia’s Flagship RTX 3090 GPU – Everything running on a Widely Available V80 FPGA Chip Advanced Micro Devices (AMD), another major US semiconductor company.
The innovation won the highest honor at the prestigious FPGA 2025 conference, which ended on March 1st. The victory marks the first time a mainland Chinese team has won the event’s Best Paper Award, indicating that it will optimize earthquake shifts in global races. AI Hardware.
Developed by scientists at infinigence-ai, a startup based in Shanghai Jiaoton University, Tinshua University and Beijing, the model can redefine how to deploy cost-effective, energy-efficient AI systems from robot control to robot control. Self-driving cars.
An FPGA or field programmable gate array is a programmable semiconductor device that allows for post-manufacturing changes to circuits and functions. In contrast, traditional Tip Once manufactured, CPUs and central processing units, GPUs or graphics processing units, ASICs or application-specific integrated circuits have fixed functions.
FPGAs and Nvidia GPUs each have distinct advantages in video generation and general computing. FPGAs provide an architecture that can be customized for a particular application, providing greater energy efficiency and lower latency. Meanwhile, Nvidia GPUs, known for their large-scale parallel computing power, are excellent at processing large-scale data and complex computational tasks.
Based on previous research, the Chinese team developed FlightVGM, the first video-generating AI model to be trained to FPGA. Through innovation in data architecture and scheduling methods, FlightVGM has achieved computational performance that could outperform GPUs.
