Comparing the processor development model of the past few years, NVIDIA CEO Jensen Huang, in his GTC 2017 keynote, cited statistical analysis from Stanford University in March of this year, showing that while the amount of data continues to explode each year, the growth rate of traditional processors has slowed relatively, gradually deviating from the previous Moore's Law growth model. However, with the continued advancement of GPU technology, the number of transistors continues to increase, and overall power consumption continues to decrease, GPU-accelerated computing models are growing at a rate of approximately 1.5 times per year, and are expected to achieve a growth rate of up to 2025 times by 1000.
According to data released by NVIDIA, there are now more and more computing models that are accelerated by GPUs, and they are widely used in the development of deep learning, computer vision, artificial intelligence technology, etc. At the same time, compared with traditional computing models that rely on a large number of processors, they can bring higher computing efficiency and reduce the overall power consumption and size of computing devices. Therefore, more and more companies and development teams are beginning to use GPU acceleration as their main computing mode.
而在今年度的GTC 2017的參展人數達7000人，在近5年內約呈現3倍成長，而GPU技術開發人員數量更在近5年內更呈現11倍成長，目前已經累積達51.1萬人規模，同時CUDA開發工具在去年更累積下載超過100萬次。
In recent years, NVIDIA has focused on applying GPUs to accelerate deep learning, thereby driving various AI applications such as image recognition, autonomous driving, and image processing. This time, NVIDIA has further applied AI technology to IRAY Tracing to improve the efficiency of real-light and shadow rendering and further reduce noise generation, making the actual rendered images more realistic.
In addition, the benefits of GPU acceleration also allow Amazon to make its Alexa digital assistant service, which is supported by its cloud platform, more responsive. It also allows SAP solutions to analyze the proportion of specific advertisements in large numbers of videos, or to assist in the classification of various types of files to avoid possible errors caused by human intervention.
