Looking at high performance computing hardware (Figure 3, right), the situation is somewhat different, but in the end still the same. If we look at the growth in single processor performance using vector processing technology (based on the Cray T94), it is not that impressive—less than an order of magnitude per decade. This is based on performance solving a 1000 x 1000 linear system (the LINPACK-1000 "Toward Peak Performance" benchmark). If we allow for multiple vector processors (based on the Cray T932), we can do much better. However, to really see the growth here we need to compare state of the art then (Cray 1-S) with state of the art now (Intel Sandia Red) and look at performance on the largest problem solved (the R_max benchmark). R_max for the Cray 1-S is little different than its performance on the LINPACK-1000 benchmark, so again we see the two orders of magnitude per decade growth rate, which, if anything, in recent years has actually been accelerating, due to the use of massively parallel architectures.