Hi. I am trying to split data for some 15 countries. First step is to combine the country data (with osmconvert) to one large o5m file and then try to split the combined file. The o5m file is about 14 GB in size. When running splitter i give java -Xmx1600m options to splitter are: --max-nodes=1500000 --no-trim Splitter calculates around 640 areas/tiles to be created. If i understand correctly, ways are parsed completely and then starts relations. After 100 000 relations i get: ***** Full GC ***** and soon thereafter: Exception in thread "main" Exception in thread "worker-0" java.lang.OutOfMemoryError: GC overhead limit exceeded... and many more lines. After this, one thread (out of two) dies but the other seems to hang doing nothing... and i have to manually kill it later. Now, i can give java a little bit more memory to java, say up to 1800m. setting --max-nodes= larger should use less memory says documentation. With a larger max-nodes, will those tiles be faster or slower to display on a GPS unit? What is a reasonable value, 10 000 000 nodes or 2 000 000? Would --max-areas help my situation? If i use --max-areas=350 (approx half of total expected tiles), based on your experience would it be enough, or shall i better try with a value of 220 (~1/3 of the total tiles)? Regards Karl