Performance assessment on the Mark IIIfp has been published in [Hipes:90a]. In brief, a small but otherwise typical case was run both on the Mark IIIfp and on one processor of a CRAY Y-MP. Performance on 32 nodes of the Mark IIIfp surpassed that of the sequential code on the Y-MP; on 64 nodes, the performance was approximately three times higher than on the CRAY. Considering the small size of the test case, a reasonable parallel efficiency (60% on 64 nodes) was observed.