Rendering and performance engineering for games, GPU research, and Evolve, our GPU and AI benchmark. Licensed on every major console platform, with team credits earned through our Embark engagement.
Our engineers earned these credits through Traverse’s five-year engagement with Embark Studios. We bring that experience to game porting, platform engineering and co-development.
From Imagination Technologies micro-benchmarks to Embark platform engineering and Samsung’s mobile graphics work. Public projects and publications are linked above.
Years of neural and graphics expertise distilled in a benchmark
Evolve turns our benchmarking research into software you can license and run yourself. Compute, ray tracing, rasterization, drivers, energy, acceleration structures and work graphs, each workload configurable and repeatable.
I grew up in Breda, left for Stockholm to ship Battlefield and Mirror’s Edge at DICE, and came back. In 2019 I founded Traverse here to support GPU teams from early product development and pre-silicon research through every stage of silicon bring-up.
Today we’re 19 people, including 15 engineers, in the same city I started in. We build Evolve, the benchmark Qualcomm and Samsung run in-house, and we’ve done pre-silicon research for most of the hardware vendors on this page. Now we’re bringing that head start back to the studios we came from.
The mission hasn’t changed since day one: make Breda a game development powerhouse.
Learned lighting, built into the renderer. Our mobile neural radiance caching implementation was published with Samsung at SIGGRAPH Asia 2025, connecting rendering research with hardware-aware performance modelling.
Learned texture representations, from training to GPU decoding. We evaluate reconstruction quality, memory footprint and runtime cost alongside conventional texture-compression tools.
A material and rendering capture from Breda, the framework used for our texture research.
Traverse Research
GPU inference
Neural workloads inside the frame. We integrate operators, tensors, inference and training with the render graph, connecting Python model development and ONNX interchange to GPU execution.
Breda framework tooling, shared by rendering and neural workloads.
Traverse Research
Neural materials
Learned material appearance, evaluated in a working renderer. NeuBTF, published at SIGGRAPH Asia 2022, explores neural compression and rendering of bidirectional texture functions.
NeuBTF material-representation research, published by Traverse.
Platforms & graphics technology
If it has a GPU, we ship on it.
From console and mobile to desktop and handheld. We work across the platforms you ship on and the graphics APIs behind the frame.
Investigating the quality, efficiency and memory trade-offs of neural bidirectional texture functions.
Traverse × Samsung / Mobile ray tracing
A public look at our work with Samsung.
Samsung shared this Xclipse GPU demonstration on its official channel. Built in our Breda rendering framework, it shows our mobile ray-tracing collaboration in action.
BVH construction, refit and driver integration for a desktop GPU vendor, evaluated in shipped game workloads rather than only isolated synthetic tests.
Working with a mobile silicon vendor to investigate rendering performance through Evolve and neural radiance caching on their GPU.
From collaborators and benchmark users.
“I really like the benchmark, the efficiency result is very nice for tweaking the GPU”
“Overall, love the benchmark and the new technologies it uses. Keep it up!”
“I’ve worked with Jasper and his team on four different occasions: Frostbite, SEED, Embark and GPC. I’d hire them again.”
“Traverse Research were a pleasure to work with. The custom Vulkan hardware ray tracing solution they built into our Linux-optimized Unreal Engine branch is still in use years later”
“Traverse Research has been a valued partner to us for over three years; deeply technical, consistently responsive, and genuinely invested in solving the hard problems alongside our team”
MakeReviewRepeat
Good work starts with a shared build.
An initial 16-week commitment, including ramp-up. Four-week sprints, build reviews every two weeks, and direct access to the engineers in your pipeline.
In this blog post, I’ll write about how I recreated the OIDN U-Net DNN using Rust and HLSL compute shaders, and about my time as an Intern at Traverse Research.