What Will the Next 35 Years of IBM i Hardware Bring?

July 26, 2023 Alex Woodie

Will IBM move beyond 64-bit memory addressing for the IBM i platform? What about the potential for quantum processors in the next three-and-a-half decades? These were some of the ideas floated by retired IBM Chief Architect Frank Soltis during the recent celebration of IBM i’s 35^th birthday.

One of the reasons the IBM i platform is still kicking 35 years after the launch of the AS/400 back in the summer of 1988 is its capability to incorporate new hardware without throwing all the software developed for it into disarray. There are limits to this adaptability, of course, but it’s a strength that has served the IBM i platform and its customers well over decades of use.

Soltis likened this characteristic to a superpower during the North American call celebrating the AS/400’s birthday on June 21.

“From our standpoint, what makes the AS/400 and IBM i invincible in a sense is . . . technology independence,” he said. “The hardware platform is independent of the upper-level technology. And that allows us to slip out the old and insert the new. And it’s, for all practical purposes, seamless.”

The failure to adopt new hardware is what doomed the AS/400’s closest competitor in the minicomputer business, Hewlett-Packard’s HP3000, Soltis said.

“The problem they had is inability to move forward in technology,” he said. “So as they tried to move their platforms to 64 bit technology, they very much destroyed the software that they had associated with it, which is something you just simply can’t do for your installed base. So the secret, if you will, is to just walk away and say ‘Well we really weren’t interested in that particular part of the business.’”

The AS/400, of course, moved from a 48-bit architecture, which it had at the launch in 1988, to a 64-bit architecture several years after it was initially launched. Amazingly, the platforms’ Technology Independent Machine Interface (TIMI) layer enabled the bulk of existing applications to run unchanged as it moved from 48-bit CISC processors to new 64-bit RISC processors.

That may have been the biggest hardware change in the history of the AS/400, but it certainly wasn’t the only one, Soltis said.

“Somebody pointed out to me that there have been over 20 different hardware platforms under the AS/400 and IBM i over the years,” he said during the Zoom call. “We continuously changed the hardware platform. And aside from more performance, more capacity, better operating, nobody notices the difference. You can run the oldest software – not that we encourage that necessarily! – but you can run some very software on the newest system out there. No one else can do that.”

And that move from 48-bits to 64-bits may not prove to be the biggest hardware change yet, depending on how IBM decides to flex its hardware independence muscle in the future.

“My reasoning is the platform itself has all of the characteristics to incorporate the latest technologies as they come along,” Soltis said. “As we know, new technologies continue to pop up. The question is can you incorporate them?”

Before the AS/400 moved to 64-bit processors, there was a plan to consolidate all of IBM’s computers on a single processor. According to Soltis, that put it on a path to adopt a 32-bit processors. Luckily, he helped steer the platform another direction.

“We realized that, after the introduction of the AS/400, we needed to go to a full 64-bit implementation,” he said. “And of course, our engineers in Rochester were more than happy to put together their own form of the 64-bit computer system. Well, it turned out that at the time, the executives at IBM we’re looking to have a common platform for all the systems.”

As Soltis told the story, the president of IBM actually came to the lab in Rochester, Minnesota, which was unusual, because the lab was so remote. In Rochester, the IBM president explained that the future of the platform would be the Power processor.

“And we looked at it and said ‘This is a 32-bit engineering, scientific processor. Not too many people want binary floating-point operations for their business,” Soltis said. “It just didn’t have most of the things that you needed. So the first reaction was, no, we don’t want to do that.”

Soltis sensed that telling the president of IBM to keep his 32-bit Power processors away from his AS/400 platform would be “a career-ending move.” But he did it anyway, and insisted that the AS/400 have a suitable processor.

“Like I said, he was an engineer,” Soltis said of the IBM president. “He agreed that . . . the technology that we would need for business computing, we would get. And he instructed the head of research – research owned the Power architecture, I believe still does – and he directed them to give us whatever we needed.”

That would prove to be a decisive moment in the AS/400’s hardware legacy. From that moment on, Power would be the main processor for AS/400 applications (aside from all the outboard processors used for I/O, encryption, etc.).

With more influence over the design of the Power chip, Soltis and company sought to build some insurance into the hardware.

“So there’s several things in that Power architecture today — obviously single-level store was a major piece of it – but we were also concerned about a 64-bit address being not enough so we managed to get in 72- and 80-bits of address into the Power architecture,” he said. “So if you think about how big a system you can get with 64-bits, now multiply that by, you know, 2 to the 16^th. So we haven’t implemented it yet, but clearly that’s there.”

Convincing the IBM decision-makers to move in a certain direction isn’t always easy. It’s not enough to be an engineer, Soltis said – you have to be a salesperson too if you want an IBM executive to greenlight your idea.

“With the development of the S/38, we had many, many times where that program would have been killed because people looking at it said ‘You know, it won’t work, it’s too different. IBM doesn’t build computers that way.’ So you had to first of all be confident that you were right in what you were doing and you had to be able to explain why you were correct in what you were doing. And so we were able to do that and get most of what we wanted. In fact, I believe we got everything we wanted for Power, even if it hasn’t been used yet?

Soltis has been watching some of the technologies that could someday replace silicon in processors. IBM has quantum computing products, and that poses a fascinating question.

“Could we move to a quantum computing base?” Soltis asked. “And my answer is yes, probably. So I think the future is very bright for the next 35 years.”

RELATED STORIES

Dr. Frank Reminisces on 35th Anniversary, Looks to Platform’s Future