There might be different errors cropping up with different machines, but there should hardly be any difference between HiSeqs 2000 and HiSeq2500. This Blog from Illumina also claims high concordance of the data: https://blog.basespace.illumina.com/2012/11/26/high-concordance-between-hiseq-2000-and-hiseq-2500-data/
I would be more concerned to compare HiSeq and NextSeq/NovaSeq data since the sequencing readout is different, or Illumina and Ion Torrent machines where the whole process is different