Windows NT 4.0: Released to Manufacturing July 31, 1996

It was 29 years ago this week, on July 31, 1996, that Microsoft released Windows NT 4.0 to manufacturing. The workstation version appeared on store shelves a month later, with the server version following in September. In many ways, version 4 was a breakthrough release for the technology. Let’s take a moment to look back at what made Windows NT 4.0 such a breakthrough.

Windows NT 4.0 advantages over earlier versions

The most visible change for Windows NT 4.0 was the user interface. Gone was the Windows 3.1-style program manager and file manager, replaced with the Windows 95 Explorer interface.

The version also included new functionality, including Group Policy. We still use Group Policy extensively in Windows today.

And one change, easily forgotten today, regarded graphics drivers. Microsoft moved the video subsystem from user space, ring 3, to kernel space, ring 0. There was a lot of talk about Ring 0 versus ring 3 on July 19, 2024 thanks to the large computer outage on that day. In 1996, this move was controversial, for the same reasons. The fear was that a malfunction in the graphics driver would now be able to take down the entire system. But the trade-off was much improved performance. It meant Windows NT 4.0 could be used for serious graphics work.

It even meant Windows NT 4.0 wasn’t a complete nonstarter for gaming. While many Windows games didn’t run under NT, the ones that did performed rather well. NT4 was never a big gamer OS, but some gamers unironically loved and recommended it, if their games of choice were among the titles that happened to work.

The secrets behind the breakthrough

I do think some of the reason for NT4’s success was due to things beyond its control. It was an improvement over version 3.51, but 3.51 had a reputation for being a bigger improvement over its predecessors than 4.0 was. Windows 2000 was a bigger improvement than either of them, bringing better stability, plug and play, and even more new functionality.

Memory

Memory had something to do with it. It’s not that NT4 was more efficient with memory than earlier versions. When it came to memory, Windows NT could never get enough. Well, until you gave it the full 4 gigabytes it was capable of using, but nobody was doing that in 1996.

The problem for Windows NT always was it needed 16 MB of RAM at minimum, and you would be a lot happier with 32 or more. I remember feeling very limited even with 32. Problem was, memory was expensive. I remember selling computers at retail in 1994, and a 4 MB module cost $150. We ran a couple of promotions offering memory at a discounted price when it was bundled with something else, but outside of that, memory was stubbornly stuck at $150 all year that year. That meant 16 MB cost $600.

Everybody was talking about Windows NT in those days, but I met exactly one person who was using it. And it was a problem for him, because he couldn’t find device drivers for any of the cool stuff everyone else had. Limited demand meant limited support.

Memory prices crashed in 1996, and the general trend held the full year. That meant when NT4 hit store shelves that summer, and magazines like PC/Computing proclaimed it was better than Win95 and you would never have to reboot again, it was feasible to try it out. The promise of never, ever needing to reboot turned out to be severely overblown, but it was noticeably more stable than Win 95. Quantifying the difference ended up being one my first IT analysis projects.

CPU power

Intel even had a fancy new CPU that liked Windows NT 4.0 a lot. That was the Pentium Pro. NT could run on RISC CPUs too, including MIPS, Alpha, and PowerPC. But most software was written for Intel, so having a high-performance Intel CPU helped. The Pentium Pro was a disappointment when running Windows 95, but it ran Windows NT 4 really well.

How the Windows NT4 migration paid for itself

If your fledgling IT department had some kind of a help desk ticketing system, it was possible to quantify NT4’s impact. You could compare the number of tickets people using each operating system opened, and attach a dollar figure to the downtime. I worked on the justification for migrating to NT4 at two different employers in the late 90s.

The value of the recovered productivity more than covered the cost of a memory upgrade and a Windows NT license. In some cases, the value of the recovered productivity could justify buying a brand new pre-built computer with Windows NT pre-installed on it, and then there wouldn’t be any question of compatibility issues.

Windows 2000 was a bigger hit, and XP/2003 were even bigger still, if only because that was the generation when Microsoft had one code base that with a few tweaks could serve everything from the casual home user to a multinational enterprise. But they all built off the momentum from NT4.

How Windows NT 4 survived as legacy IT for decades

Rumor has it you can still find Windows NT4 lurking in forgotten corners of large businesses even today. It’s been a few years since I have been in a position to find these systems, and I do think we overestimate how numerous they really are. But some NT4 systems did last long beyond their use-by date, and I know exactly how it happened. Some of them were even partially my fault.

Circa 2005, I migrated a few NT4 systems to VMware because the hardware they were running on was aging badly and starting to fail. We couldn’t rebuild those systems because the software they were running was business critical at the time, nobody knew where the original installation media was, and the software vendor was out of business. The problem with that was by migrating those systems, I eliminated every incentive to find viable replacements for them. It turned into something they could kick the can down the road for years, if not decades. After all, they could just load the image from one VMware server to the next as they upgraded their VMware infrastructure. I warned them that’s what would happen, but they told me to do it anyway.

My story wasn’t an isolated incident

At several stops in my it career over the next decade, I uncovered similar systems, and they always had a backstory. In some cases, someone even told me the name of the former co-worker who originally built that server and presumably discarded the original media. Not that there was anything any of us could do about it, but sometimes people even remembered names.

Those Windows NT systems that survived a decade or two beyond their official end of life in 2004 are the exception. Even though companies built up sizeable Windows NT4 fleets rather quickly, they moved on to the subsequent operating systems much more quickly than they do today. The benefits of moving on included things we take completely for granted today, including USB support, Plug and Play, and Active Directory. Migrating to new operating systems was different then. Once end users recognized the advantages of NT4, they were happy to migrate to it. When the IT department I worked at from 1998 to 2005 dragged their feet on migrating to Windows 2000 and 2003/XP, end users complained. Loudly. It was a big contrast to the migration to Windows 7, Windows 10, and Windows 11.

One piece of Windows NT 4.0 legacy you can’t escape

But in the course of studying Active Directory security as a part of my current job, I rediscovered another relic of Windows NT4 that still exists even in the very newest version of Active Directory. That is something called the PDCe. PDCe stands for Primary Domain Controller Emulator. Back when Windows 2000 was still in beta, the big promise was that you wouldn’t have to worry about primary and secondary domain controllers anymore, that in Active Directory, all domain controllers are equal. Active Directory didn’t completely deliver on that promise. It’s true that usually you can ignore the difference and treat all of your domain controllers as equal, but that primary domain controller emulator is the most important of them.

If you ever run across the PDCe, that is the Active Directory domain controller that emulates the functions of the old Windows NT4 primary domain controller. If you are like me, the day you first learned about that, you weren’t having the best day of your career. At least I wasn’t.

Dave Farquhar

David Farquhar is a computer security professional, entrepreneur, and author. He has written professionally about computers since 1991, so he was writing about retro computers when they were still new. He has been working in IT professionally since 1994 and has specialized in vulnerability management since 2013. He holds Security+ and CISSP certifications. Today he blogs five times a week, mostly about retro computers and retro gaming covering the time period from 1975 to 2000.