Obbe Vermeij says GTA 3 battled crashes, poor performance and unfinished areas until its final six months, when the team pulled the game together.
Grand Theft Auto 3 became a landmark open-world game, but its development was far from assured. Rockstar veteran Obbe Vermeij recalls a project beset by crashes, low frame rates and gaps in the map, with the team only bringing it together during the final six months. That account complicates the familiar story of a hit that seemed destined to reshape its genre: the finished game’s confidence and scale came out of a production that, for much of its run, still had basic technical problems to solve.
Vermeij’s recollections also point to how much of the game’s identity was formed through trial and revision. The team was building a new kind of 3D experience for the series, while its plans for missions and story were still changing. There wasn’t a settled blueprint waiting to be executed. The late push to make the game run, and the willingness to keep testing ideas, were both part of how it reached release.
A late push to reach 30 frames per second
For much of production, the game struggled to run reliably. Vermeij says the frame rate looked likely to settle at 20 frames per second, and the map still had holes that could send players falling through the world. In an open-world game, those problems cut into more than visual polish: inconsistent performance can make movement feel unreliable, while gaps in the map undermine the basic promise that players can explore its spaces. The closing months therefore involved getting the underlying experience to hold together, not simply adding a final layer of shine.
“It crashed frequently, the frame rate was terrible, and the map was full of holes,” he wrote. “Only in the final months did we manage to hit 30.”
Raising performance to 30 FPS gave the team a more dependable target, but Vermeij’s description makes clear how much work remained before that point. A project can look ambitious on paper and still be fragile in play; crashes and holes in the world expose that gap immediately. The numbers here are underwhelming by the standards players might expect from a finished release, but they make the final turnaround more striking. Getting there only in the closing stretch left little room for the team to treat technical stability as a settled problem.
Two developers started the project
The work began on a small scale at DMA Design, before the studio took the Rockstar Games name. Leslie Benzies and Aaron Garbut were developing GTA 3 while other staff worked on a planned isometric game set in Miami. That project was canceled, and Take-Two Interactive, DMA’s new parent company, approved the 3D game. The change in direction meant the team had to establish what a 3D Grand Theft Auto could be, rather than carry over a finished design from the planned game.
The team rebuilt the game from scratch. Vermeij says just two people had worked on both GTA 2 and GTA 3, and the new title reused neither code nor artwork from its predecessor. That reset likely made the work harder: familiar names and ideas could carry over, but the production itself had to find new technical and creative answers. In this category of project, changing the format affects more than how a game looks; it changes how players move through the world and how missions can use it.
Mission ideas took shape through experimentation: developers proposed them, the team tried them, and the fun ones stayed. That process sounds informal, but it also gave the team a practical way to judge ideas in a game whose structure was still taking shape. The story came later. Catalina’s attempt to kill Claude, which sets the opening plot in motion, was added well into development, suggesting the narrative was fitted around a project that had already begun to find its form through play.
Signs of the changing design remain in Liberty City
Early plans shifted as Benzies moved missions between characters, Vermeij says. The structure became clearer after voice recording wrapped, but traces of the revisions remain in the finished game. Some locations have more detail than their role might suggest because a mission contact was once placed there before being moved elsewhere. Those spaces preserve the residue of decisions players may never see: a location’s apparent purpose in the final game doesn’t always explain why it received the attention it did.
That kind of leftover detail is a useful reminder that game worlds aren’t always designed in a straight line from map to mission. A space can be built for one use, then retain its shape and finish after the plan changes. (And this is the part most people overlook.) The world players explore is also a record of production choices, including the ones that were revised before release.
Confidence in the game’s prospects was also limited. Vermeij recalls a disappointing E3 showing in May 2001, where some people showed little interest and others considered the game too violent. The launch was “OK,” he said, but sales kept going rather than fading after release. That distinction matters: an initially modest response didn’t settle the game’s longer-term standing. Its audience continued to find it, even though the early reaction gave the team little reason to assume it would become a defining success.
A small team, a lasting mark
GTA 3 went on to shape the market for open-world games, with Rockstar following it with Vice City and San Andreas, while other studios pursued the format through series such as True Crime. Its influence can be understood in part through that response: the game helped make the open-world format a more visible direction for major releases. That doesn’t mean every later game copied its approach, or that its success was obvious during development. The contrast between its troubled production and its later influence is precisely what makes Vermeij’s account useful.
This is more significant than it looks. If you’re working in this space, the story is a reminder that a genre-defining release can emerge from a production that hasn’t found its footing until late. Yet the account doesn’t reduce the result to a technical rescue: the team also changed missions, refined the story and made a city whose locations carried traces of their earlier purposes. A late performance improvement could make the game playable; it couldn’t, by itself, explain why the finished work stayed with players.
For Vermeij, the strongest memory is the close-knit team: many developers were young, had recently moved to Edinburgh, and spent time together outside work. He remembers “a group of friends who happened to be working on a project together.” The recollection offers a more personal measure of the project’s scale. A small team, a difficult production, and a game that kept finding an audience.
Not a smooth climb.
Discussion
Sign in to join the discussion.