The problem
Traverse Research develops and investigates GPU workloads for hardware evaluation. We connect repeatable measurements to rendering and architecture questions so a result can inform a design decision.
Hardware behaves differently under different workloads. A change can help one stage while moving cost into another. Understanding that trade-off requires control over the workload, its settings and the measurement conditions. Repeatability is what makes comparison useful.
What we do
Our work spans early and pre-silicon investigation as well as consumer hardware. That engineering is also a product: Evolve, our cross-platform GPU and AI benchmark, launched in March 2025 and run in-house by Qualcomm and Samsung. Its features, downloads and business offerings live on its own website.
For a research engagement, we agree the question first: which resources matter, what comparison is meaningful and how the evidence will be used.
A benchmark other teams can evaluate
Evolve launched in March 2025. In February 2026, LTT Labs published an evaluation of Evolve 13.1 across five Android devices, with repeated measurements and comparisons against other GPU benchmarks. Its results and methodology provide an independent view of the product built on this engineering.
Read the LTT Labs evaluation · Explore the Evolve case study.