Mariner 9 used magnetic core memory. It had the awesome amount of 512 words (quadrupled from the original design as requirements grew).
The document never actually comes out and says so, but it's clear from this description of a problem they had:
The memory vacuum problem became evident the first time the
memory was subjected to vacuum as part of the required TA test.
vacuum sensitivity, which resulted in catastrophic failure of the magnetic-core memory plane, was traced to trapped air underneath the polyurethane-coated memory plane.
Under vacuum this trapped air expanded underneath
the plane such that the very fine magnet wire that is strung through the
magnetic cores was stretched and broken.
The solution to the problem was,
of course, the elimination of the trapped air.
This was accomplished by
more careful application of a two-sided adhesive tape used to hold the
magnetic-core plane to the mounting board, and better application of the
polyurethane coating material.
Development and Testing of the Central Computer
and Sequencer for the Mariner Mars 1971
Information on the computer can be found in the design document referenced in the one linked above.
The design selected was a programmable sequencer, with a 128-word core memory.This programmable sequencer, with a memory which could be updated or modified by ground command...
The other subsystem capable of commanding spacecraft events is the CC&S. Figure 9 shows the CC&S
with its functional interfaces. Basically, the CC&S is a
special purpose computer which has extreme flexibility
and can be reprogrammed in flight. Its primary purpose
is to provide event actuation at certain times which are
specified prior to launch.
CC&S = central computer and sequencer
This document also reveals that the word length was 22 bits.
Mariner Mars 1969. Volume 1 - Development, design, and test Final project report