PCA is a way to visualize high dimension data by projecting onto 2 or 3 planes that are chosen by the variance of data. When using PCA to visualize a gene expression table, you can typically get some insight into natural groupings, however it is important to first remove low expression gene IDs and normalize data. TO normalize, you can use log-transform and sometimes, Quintile Normalization. QN is especially popular when dealing with microarray data.
Here are some videos of how to perform PCA analysis on the T-BioInfo platform: