This paper uses a simple modification of classic Kohonen network (SOM), which allows parallel processing of input data vectors or partitioning the problem in case of insufficient resources (memory, disc space, etc.) to process all input vectors at once. The algorithm has been implemented to meet a specification of modern multicore graphics processors to achieve massive parallelism. The algorithm pre-selects potential centroids of data clusters and uses them as weight vectors in the final SOM network.In this paper, the algorithm is used on a well-known KDD Cup 1999 intruders dataset.