In theory at least, XMLParser
(a SAX implementation) can be more efficient (speed-wise, but especially memory-wise) than SimpleXML
(a DOM implementation), particularly when dealing with larger documents (DOM implementations load the entire XML tree into memory, which can be taxing on resources, whereas SAX implementations allow the programmer to simply traverse the XML without necessarily storing anything in memory, leaving entirely up to you what, if anything, you want to retain in memory) -- see benchmarks comparing various DOM to various SAX parser implementations.
I say can because performing SAX traversal is programatically more complex than traversing or querying the DOM, and a poor use of SAX (or use of SAX in situations where DOM would be better suited anyway, e.g. if at the end of your SAX traversal you ended up loading yourself virtually the entire tree in memory) can actually result in poorer performance than the use of a DOM API.
Cheers,
V.