It seems that you have genome wide methylation data measured using Illumina bead chips, and your file has been generated as a genome studio final report, correct? In this case I'd recommend considering the following points:

1. You may be better off not starting from the report file but from the idat files instead (the original files that went into the genome studio project) and use the appropriate R-packages (methylumi, minfi, etc.) from the Bioconductor repository for processing the data
2. Of course this requires sufficient memory. So the machine should have at least 8 GB RAM available (better 16)
3. There are some packages that claim to be able to load genome studio reports (methylumi). But I'm not sure about that.
4. If you still want to split the file using perl, the script should work for any number of columns per individual. Thus avoid hard coding this number in the file, as you may discover that you are missing some information and you have to regenerate the report with different settings and a different number of columns per subject.
5. Unfortunately many available programs for data processing (in particular some R functions in several packages) are not capable of obeying i18n rules and thus just use "." as decimal separator. Or they use the encoding in inconsistent manner. To ensure data integrity it seems best to change your language settings to English and as well to change all "," in the data file to ".". So one of the 1st things to do for each line after reading in would be "tr/,/./;", if you are using "," as regular decimal separator - like in your provided data file.
6. Under the precondition of enough available file handles it is possible to generate all output files in one pass (Just as poj's solution assumes; e.g. for newer windows platforms this number is practically unlimited (nominally about 16 mio per process). If this is not the case you'd have to do this in several passes.

Does poj's suggestion suit your needs?