What I meant is duplication of read ached mechanism adopted by both file system and DB_FILE_MULTIBLOCK_READ_COUNT.
The read mechanism available under file system already caches some data in the file system memory, thinking that it can serve to the next request very quickly without reading from the disk, but in reality it is a waste in a OLTP environment where the request for new blocks are at random. Even in DW environment this wastage is very high. Second, by setting the DB_FILE_MULTIBLOCK_READ_COUNT to a higher value, the OS reads more unwanted data than what is required. In both cases, the assumption is subsequent requests may be served if the data requested is available in the cache. That is why first time the user process will wait longer.