@AmpBenzScientist different tools for different jobs. For highly independent data then an Aiyara Cluster or any sort of distributed cluster design will get you the most bang for you buch.
For interdependent data on the other hand it will be useless. However may system may still knock it out of the ballpark in that situation if the data/algorithms are of the right nature (lots of the same computation done across interdependent data where the output of one computation may be the input to the next).