Over the last few weeks we've checked in with Jeff Kell, the system manager at the University of Tennessee at Chattanooga. The university powered off its last two HP 3000s not long ago, and along the way has mounted dozens of Unix and Linux CPUs and virtual servers to replace that pair of MPE machines. We asked him what he believed the school's IT group had spent on MPE over 37 years -- and limited the question to the capital costs of systems. (Ownership cost is much harder to calculate across four decades.)

Kell, who founded the HP 3000 listserve and newsgroup, as well as chaired the SIGSYSMAN group for Interex over the years, said "We have had comparable expenses with each iteration of the 3000's life-cycle." Across those decades, the university owned Classic HP 3000s based on CISC technology, then early PA-RISC servers -- new enough in that generation to be considered "Spectrum" 3000s -- then later-model PA-RISC units, and finally the ultimate generation of HP 3000 hardware.

"In short, it was an expenditure in the low six figures, once every decade," Kell said. 

We ran Series II, then Series IIIs, and the tags were low six-figures in the 1970s. We then got some 950s in the late 1980s (we had some early Series 950 deliveries) at about the same price point. Then the 969 in the 1990s, again about the same. And finally, the A/N-Class during this century.

Comparisons to two points seem worthy. The pricing for the value of high-end 3000 computing remained constant; at the time of the late 1980s, for example, a Series 950 was the most powerful 3000 available. Then there's the comparison to the expenditure of acquiring the hardware to support dozens of servers, virtual and otherwise. The low six figures won't buy much toward the high end of business critical computing gear over a decade, using today's commodity pricing. The newest servers might seem cheaper, but they don't give durable service for 10 years per installation, like the ones at Kell's shop did.