first of all it does matter, because this model obsoletes the 520 series and hence will end up in enterprise DCs and workstations. the 335 equivalents (with lower binned chips) are the ones that will end up in ultrabooks. hence, the fact that they believe that power to performance is a more important global design goal is an interesting and important development. power IS important in DC distributed applications, but IOPs are more important in monolithic DC applications. the fact that they have globally targeted what look like distributed applications for their enterprise line is quite interesting. it may be a strategic engineering choice or it may be due to the specific yields and characteristics of their 20nm flash dies. if it's the latter, that's extremely interesting information.
also i mention fusionIO because fusionIO is currently dominating the monolithic server applications market, even though they are fabless. it's possible that intel simply looked at their core competency (their fabs) and decided their best opportunity was optimizing everything for power per IOP and to mostly cede the monolithic market to fusionIO (for full units anyway -- i believe the intel/micron 20nm chips can be purchased by OEMs, and the controller, at its base, is just an SF2281 die with some bug fixes).
eta: the lossless compression algorithm (basic huffman blah blah) is implemented at the hardware level in the sandforce controllers (all of them). the basic idea is that they increase bandwidth by making an assumption that most data is compressible (this is often safe to assume in distributed enterprise sitations) and then compressing as they write data and decompressing as they read. because the limiting factor is latency of an actual write op to the chip, you gain severa
