Intel’s power a lure for legacy engines

Work is well underway to get OpenVMS to the Intel x86 line. Hewlett Packard has no part of that work, a project going up in the VSI labs. But Hewlett Packard Enterprise added x86 blades to its Superdome servers more than 10 years ago. Even while the Itanium chips still appeared to have a future, Hewlett Packard could see a bigger one. A bigger world, with better performance, surrounds the Intel designs.

It’s not that the Intel Xeon-x86 line has a superior architecture. The AMD processors for the Windows market perform better. EPYC processors out-perform the Intel designs. But this advantage enjoyed by x86 is all about market size. It’s similar to the one that Android holds in the mobile OS markets. One OS, but many implementations from a wave of vendors, puts Android into the lead.

Why should a legacy OS owner or administrator care? As it turns out, x86 is more than the past. It’s the future of computing, too. Hewlett Packard customers were told differently when they invested in other architectures, like PA-RISC, or Integrity, or even Alpha. Those were going to be the successors to x86. How could something like x86 even survive into the 21st Century, the legacy users were asked. Something newer was going to be enough better to overtake x86.

Except those things were better in just one way. They were newer, yes, perhaps better engineered, and so they had the full faith of the legacy vendors behind them. Those architectures didn’t have the faith of the market behind them, though. The market faith was behind x86. The processor survived and morphed into Xeon and retained customers, in part because the x86 creators, Intel, had a good thing going. Business was good, even while the newcomers looked better.

So as a legacy customer, how do you participate in the x86 market? Your engines for your apps are SPARC, or PA-RISC, or Alpha, or VAX, right? As it turns out, the way forward is emulation, or in the rare case of OpenVMS, an x86 port of the OS. A mighty project indeed, that one from VSI. HP has passed up doing that x86 work for HP-UX, so there’s a vague future for that Unix. Perhaps emulation is somewhere in the future for an HP-UX processor.

Technologists will tell us that it’s unfair for better designs to run behind x86 in the computing race. They’re not incorrect. That’s a race that happens on whiteboards and PowerPoint slides, though. The race that matters is to run fast enough to stay inside the pack. That’s a pack of x86 users, bigger than any other group. Safety in numbers, and all of that.

Comments

  1. Oliver Jones

    AMD’s EPYC 2 designs also deliver memory encryption across VMs. Just in terms of security, completely ignoring the fact that EPYC 2 beats Intel on performance, this is a killer feature.

    I would not consider buying into a platform that doesn’t have at least the level of protection EPYC 2 has – especially with vulnerabilities that can bypass hypervisor protection.

Leave a Reply