I'm Having a ML problem where my data set contains 60 features labelled into 3 groups (0, 1, -1).

I want to use K-means on that data and plot it on a 2D plot to see how close data with label x to data with label y

I was thinking about using PCA and transform the data from 60D to 2D, but It only retain 60% of the variance!

1. Is this a good approach for the problem?
2. If so, does 60% suffice?
3. Are there any other/better approach for this?