Enterprise Virtualization,, AutomaticReasoning, Deep Learning and High-performance Computing
PR4904P1 is suitablefor the mainstream 32-bit and 64-bit HPC applications at present, mainly aimingat the applications in traditional HPC fields such as bio-informatics,computational chemistry, computational finance, computational fluid dynamics,computational architecture, data science, security monitoring, electronicdesign automation, perception and computer vision, machine learning, medicalimaging, numerical analysis, weather and climate, etc.
As a new application field, Deep Learningis the hot point of machine learning in recent years, and it has madebreakthrough progress in the fields of image and voice recognition. It makesthe combination of big data and deep neural network model, parallelizes thedata or deep network model in GPU cluster to accelerate the program executionefficiency. Using GPU to accelerate deep learning and train deep learningnetwork can give full play to the efficient parallel computing capability ofthousands of computing cores of GPU. In the case of massive data trainingscenarios, it is characterized by shorter time and fewer occupied servers. It hasbecome the preferred solution in this field and is widely used in the Internetindustry that using GPU cluster as infrastructure to build deep learning/machinelearning platform.
With the use of GPU in the field of highperformance computing, CPU is no longer the only choice for computing chips. Comparedwith CPU, GPU has more powerful computing power (at present, NVIDIA's latestV100 has 5,120 CUDA cores and 640 Tensor cores), with simper task processingmode. Now it is gradually applied to various fields of high-performancecomputing, helping the Internet industry to develop rapidly. With its powerfulcomputing power, GPU attracts users to use GPU to improve the executionefficiency of applications. At the same time, GPU also has the advantages oflow cost, high performance and low power consumption, which reduces the overallcost of ownership of users.