chips ahoy

An interesting interview on next gen SPARCs at Aces Hardware [via Solaris Central]. I guess it's obvious by now there is a definite trend towards on-chip multicomputing in chip designs. How much real software performance boost can we get? On the flip side, do we need some new programming techinique, e.g. new languages or upgrade of compilers, runtime systems or OS to take advantage of this performance boost? Solaris no doubt will scale just fine but what about user apps?