In the first example, two truly 3D meshes of SGI logo (Fig. 1) have been generated. The coarser one, consisting of 14064 nodes and 18966 elements, have been obtained using from 1 to 5 processors on SP2 and from 1 to 9 processors (except the numbers corresponding to unsupported configurations) on Paramid. The finer one contains 67964 nodes and 117492 elements and has been generated using from 1 to 8 processors on SP2 and from 4 to 15 processors on Paramid. Unfortunately, it was not possible to use higher number of processors because of temporary lack of resources. The speedup obtained on SP2 machine is plotted in Figure 4 and the time, speedup and efficiency is summarized in Tables 1 and 2. Similarly the results achieved on Paramid machine are displayed in Figure 7 and summarized in Tables 7 and 8. The coarser mesh of SGI logo is the only mesh which has been generated on Paramid using a single processor. The profiles of the speedup achieved on SP2 and Paramid for this coarse mesh are quite similar with the only difference that the saturation of speedup is observed somewhere around 10 processors in the case of Paramid contradictory to 6 processors in the case of SP2. This can be explained by different CPU performance of SP2 and Paramid. Since very coarse mesh is under consideration the latency time which is likely to be rather similar for both platforms is dominant to the overall communication costs. The SP2 machine seems in this particular case to be approximately 6 times faster than Paramid and therefore the communication costs on SP2 are proportionally much larger with respect to the total computational time than it is on Paramid.