Abstract: In order to solve some problems those 3D modelling visualization is facing such as massive model data, long process time, slow 3D rendering, clients depend on hardware too much, etc.This paper proposes a good solution to build a 3D visualization platform for city geology in cloud-end based on Spark big data technique, this platform consists of four layers including data storage, data dispatching, cloud-end rendering and client visualization.For managing data effectively, big geologic data is divided automatically according to policy that called “vertical layering, horizontal blocking”.Using Spark Streaming+GPU to deal with parallel computing and 3D task dispatch to perform 3D distributed rendering, centralized integration and visualization display during the process of computation and visualization for 3D modeling.Five high performance virtual computers of GeoCloud are used to build prototype of 3D visualization platform for city geology, and geologic data, borehole, high precise meshes, tilt photography cover Xiongan area are all employed through the whole testing work.The total data volume is nearly 410 GB.The test results display that the client without supporting by 3D Jet access the 3D visualization platform for city geology in GeoCloud, thirty pictures will be generated within a second and return to client application, so that user can interact with cloud-end 3D platform fluently.This new kind of “client send request, 3D computing in cloud” operation pattern will contribute much and promote city geology to become practical as normal application.