Reason-able views represent a sort of search engines, or indices that allow efficient access to parts of the Linking Open Data (LOD) cloud. Linked Data Semantic Repository (http://ldsr.ontotex.com) combines several of the central datasets of LOD (e.g. DBPedia, Freebase, Geonames) in a repository containing more than 1B explicit statements. After inference, there are 4B different retrievable statements describing about 250M resources. For some entities LDSR contains fairly big volumes of data (e.g. 1380 statements for San Francisco), thus, a good visualisation facility need to find a way to present a “good” fraction of the information. We have developed a mechanism, called RDF priming, for ranking the importance of the information in an RDF graph with respect to both its topology (e.g. PageRank) and the relevance to specific query and user context. We present the benefits of RDF priming for LDSR exploration, along with the technical solution allowing us to perform ranking in a graph of several billion statements in real time. We apply the CUDA architecture that allows using NVIDIA Graphical Processing Units (GPU) for massively parallel computations. In this context, a CUDA card priced around $400 performs several times faster than a dual-CPU server worth $4000.