When members gather from the 3000 community, they don't often surprise each other these days with news. The charm and challenge of the computer's status is its steady, static nature. We've written before about how no news is the usual news for a 40-year-old system.
But at a recent outing with 3000 friends I heard two pieces of information that qualify as news. The source of this story would rather not have his name used, but he told me, "This year we actually sold new software to 3000 sites." Any sort of sale would be notable. This one was in excess of $10,000. "They just told us they needed it," my source reported, "and we didn't need to know anything else." A support contract came along with the sale, of course.
The other news item seemed to prove we don't know everything about the potential of MPE and the attraction of the 3000 system. A company was reaching out for an estimate on making a transition to the Charon emulator. They decided not to go forward when they figured it would require $1 million in Intel-based hardware to match the performance of their HP 3000.
"How's that even possible?" I asked. This is Intel-caliber gear being speficied, and even a pricey 3000 configuration shouldn't cost more than a quarter-million dollars to replace. It didn't add up.
"Well, you know they need multiple cores to replace a 3000 CPU," my source explained. Sure, we know that. "And they had a 16-way HP 3000 they were trying to move out."
Somewhere out there in the world there's an HP 3000, installed by Hewlett-Packard, that supports 16 CPUs. Still running an application suite. The value is attractive enough that it's performing at a level twice as powerful as anything HP would admit to, even privately.
A 4-way N-Class was as big as HP would ever quote. Four 500-MHz or 750-MHz PA-8700 CPUs, with 2.25 MB on-chip cache per CPU, topped the official lineup.
Unix got higher horsepower out of the same HP servers. An 8-way version of the same N-Class box was supported on HP-UX; HP would admit such a thing was possible in the labs, and not supported in the field. But a 16-way? HP won't admit it exists today, and the customer wouldn't want to talk about it either. Sometimes things go unreported because they're too big to admit. It made me wonder how much business HP might've sustained if they'd allowed MPE to run as fast and as far as HP-UX ran, when both of those environments were hosted on the same iron.