In theory there are no significant differences other than the marking of the chip.
Electrically they function the same.
In software they function the same.
The same exact chip is used to make both.
Some say they sort them a little differently in terms of the allowed VID voltage for one versus the other, but that probably doesn't matter much since with any given single CPU you have no idea of what the VID is going to be anyway and it varies over a wide range, so for the most part the two chips span most of the same possible range of voltages coming from the factory.
Some say that they may do some firmware level tweaks to some of the chip's algorithms that can make the xeon more well 'tuned' to perform for typical server type software operating software, and the desktop model gets slightly different algorithms to tune it to work best for average desktop types of operating software. Things like memory prefetching stragegies etc. Things that don't work perfectly one way or another, it's just sort of a best guess / compromise about what is probably most often a good guess even though you're always going to guess wrong a good percentage of the time depending on the software in use. Personally I've seen no credible backup to suggest that there ARE such differences in firmware/microcode level 'tuning' between the chips. I've seen no credible claims of anyone being able to tell the two chips apart in benchmarks because of these rumored differences, though some people claim that the xeon performed 3% worse on some 'desktop/gaming' oriented benchmark of some sort, but AFAICT that is just a rumor until I see the hard data about the test methodology and repeatibility of the results they get and how often such a difference even manifests itself etc. It could just be a rumor or the result of bad testing methods or software bugs in the actual OS or program and nothing to do with the CPU.
In any case they're to a high degree of approximation equivalent.
http://processorfinder.intel.c...tails.aspx?sSpec=SLAPL
http://processorfinder.intel.c...tails.aspx?sSpec=SLAPM