Teardown Developer Breaks Down Multiplayer and Voxel Destruction Tech

Few games have captured the imagination of developers quite like Teardown, the physics-driven sandbox known for its fully destructible voxel environments. Built by Tuxedo Labs and later published by Saber Interactive, the game quickly became a standout example of how creative simulation and emergent gameplay can redefine a genre.

Instead of following traditional mission structures, Teardown challenges players to carefully plan their approach, manipulate the environment, and unleash chaos in worlds where nearly everything can be destroyed.

With the recent addition of multiplayer, the technical complexity behind the game’s systems has grown even further. We spoke with Dennis Gustafsson, founder of Tuxedo Labs and creator of Teardown, about the challenges of synchronizing destruction across players, the design philosophy behind the game’s voxel-based engine, and why building a world where everything can break introduces a unique set of technical and gameplay problems.

You can also read more about how multiplayer was added to Teardown in this devblog.

Teardown has built a strong reputation for its fully destructible environments and sandbox-style gameplay. Looking back since launch, what aspects of the game’s design or technology do you feel most proud of today?

Dennis Gustafsson, Founder of Tuxedo Labs: As a physics programmer, a fully destructible environment has been a bit of a holy grail ever since I got interested in simulation in the late nineties. To finally be able to deliver on that is probably the most significant aspect. I'm also really proud of the gameplay, which takes advantage of what the technology has to offer, using destruction not just as a visual effect, but actually useful and even necessary to complete the objectives.

How has the vision for Teardown evolved over time, particularly as the game expanded with new content, tools, and community-driven features?

Dennis Gustafsson: Teardown as a modding platform has grown much bigger than we anticipated and we're trying to lean into that. There are now mods with entire campaigns of their own with vastly different gameplay from what the game was built around. The introduction of multiplayer also expands the vision of Teardown beyond single player, with new game modes and modding capabilities we did not originally intend.

Teardown’s voxel-based world allows for incredibly detailed and dynamic destruction. Can you explain how the voxel system works under the hood and how it enables the kind of physics-driven gameplay the game is known for?

Dennis Gustafsson: Dynamic destruction in traditional games is incredibly hard, primarily because the objects are only represented as surfaces without any knowledge of what's on the inside. Add to that the difficulty with numerical precision when cutting up meshes and convexity constraints from physics engines, and you typically have to settle for prefractured objects, breaking the same way every time. Teardown objects are represented and modeled as voxel volumes on a regular grid, and the whole engine has been built around that. This simplifies a lot of things both in physics and graphics when breaking them up.

Destruction at this scale can be extremely expensive computationally. What techniques or optimizations did the team use to ensure performance remains stable?

Dennis Gustafsson: When I first started rewriting the collision detection to work with voxels, I thought it would be awfully slow, but it actually wasn't so bad as I initially thought. During early access, I spent a lot of time optimizing the physics and destruction, making use of both SIMD and multithreading to take advantage of modern hardware. For the console ports, Saber managed to improve performance further, so it's been a long journey with many people involved.

Lighting plays a huge role in Teardown’s visual identity, especially with its distinctive voxel ray tracing. How does the lighting system work technically, and what challenges did the team face when implementing real-time global illumination in a destructible world?

Dennis Gustafsson: Teardown actually doesn't implement global illumination (light does not bounce off surfaces), but we use raytracing for ambient occlusion, soft shadows and specular occlusion. The Teardown renderer is very different from most games. Like the rest of the engine, it uses voxel volumes to represent the objects, so there aren't any triangles involved. Instead, the GPU marches through these voxel volumes along the ray to see where it hits something. It sounds expensive, but we have several layers of acceleration structures in place to keep it performant. A lot of the design choices for the Teardown renderer, including the lack of global illumination, is because it was made before hardware raytracing became available.

Multiplayer introduces a completely different layer of complexity when dealing with physics and destruction. What were the biggest technical hurdles when bringing multiplayer into Teardown, particularly in terms of synchronizing destruction and physics across multiple players?

Dennis Gustafsson: We knew from the start that we couldn't synchronize destruction by sending voxel data over the network - it's simply too much data. Instead, we made parts of the network synchronization rely on determinism. The server sends specific destruction commands to all clients, which then apply them to the scene in exactly the same way and order on all machines. Some parts of the engine, like the physics, still use traditional state synchronization, so it's a combination of both approaches.

Join-in-progress was a major headache. Since destruction relies on commands being applied to all machines, such commands also need to be applied to any new client joining a game. Therefore, we queue up these events and apply them to the scene while joining. To keep things reasonable, we had to limit the size of this buffer and disable join in progress when there have been a lot of changes; it would take forever to join.

Teardown’s modding scene has become one of the most active parts of the community. How was the game architected to support modding from a technical standpoint, and how much of the original toolchain is exposed to creators?

Dennis Gustafsson: There were requests for modding support even before the initial release, so we kind of had a feeling where this was going, but our initial plan was to finish the game first and add modding support later. However, shortly after the early access release (which featured only the first half of the campaign), the community already went wild with it, reverse engineering, documenting, building their own modding tools and platforms for sharing mods. This makes us quickly shift focus and put more effort into the official modding capabilities and integrate Steam Workshop. Nowadays, we expose pretty much all of our internal tools to the modders. There are a few debug and profiling features we don't ship, mostly because they are very obscure and not really useful for modding purposes.

Have there been any mods or community creations that surprised the team by pushing the engine further than you expected?

Dennis Gustafsson: This has happened countless times over the years. I think the first time I really fell off the chair was when someone implemented a working portal gun, complete with the ability to pass objects or jump through it and even a custom raytracer to see through portals.

As the game continues to evolve, are there any technical improvements or systems you’re currently exploring that could push Teardown’s destruction or simulation even further?

Dennis Gustafsson: There has been a lot of focus on getting multiplayer ready for launch, and I think that will still be the focus for some time to come. We do have new tech we're also working on, but most of that is for a new game in a completely redesigned engine.

For developers interested in building systemic destruction or physics-driven gameplay in their own projects, what lessons from Teardown’s development would you share?

Dennis Gustafsson: I'd encourage people to not shy away from writing their own tech. It is usually advised against, but I think it's a great way to make a truly unique game. The time it takes to set everything up can be daunting, but towards the end of the project you'll have everything exactly the way you want, and can work faster than in any existing engine. Even more importantly - writing your own tech is a lot more fun!

Dennis Gustafsson, Founder of Tuxedo Labs

Interview conducted by David Jagneaux