From DedicatedStructuresVersusRdbms: "It would be interesting to study that patterns of systems and requirements where Oracle (considered the performance leader) is [allegedly] not good enough."
That was Top, right? Sigh...this digression should not be necessary.
A quick search finds e.g.:
"SAG is claiming Tamino is 10-40 times faster than Oracle 8i." http://listserv.uark.edu/scripts/wa.exe?A2=ind0002d&L=sag-l&F=&S=&P=6696
''"[Lawson non-relational database] is much faster than Oracle or the other relational databases"'' http://www.mail-archive.com/oracle-l@fatcity.com/msg71361.html
These things are not interesting. A general purpose relational database can always be beaten, speed-wise, by some special purpose piece of software.
Pay closer attention this time: this has been proven mathematically; it is not a subjective opinion. See the NoFreeLunchTheorem, like I said before.
And actually, it was obvious even before that, that this is usually true. The theorem is interesting because it proved that it is always true.