We describe the design of an embedded 128-Kb Silicoti-OnInsulator (Sol) CMOS SRAM, which is integrated alongside an array of pitch-matched processing elements to provide massively-parallel data processing within one integrated circuit. A n experimental 0.25-pm filly-depleted SO1 process was used. The design and layout of the SO1 memory core and results from calibrated circuit simulations are presented. The impact of the floating body efSect is investigated f o r both memory reads and writes. We describe threshold mismatch eflects in the sense ampli$er that result from the Jloating body voltage. Floating body effects are compared against simulated results f o r an SRAM designed in a 0.25-p m partially -depleted SO1 process.