Image
Parties involved
Treble Technologies, an Icelandic small and medium-sized enterprise, develops sound simulation tools that help companies design and improve acoustic environments and audio products. With support from IHPC, Treble gained access to the DEEP high-performance computing system at JSC in Germany. DEEP received funding from the EuroHPC Joint Undertaking. This access helped the company accelerate acoustic simulation and develop tools for virtual prototyping and synthetic audio data generation.
Challenge
Traditional acoustic modelling faced limitations in speed and accuracy, particularly for complex or large-scale environments. Wave-based simulation was too slow to be practical beyond very small spaces. Geometrical acoustic platforms struggled with accuracy at low frequencies and could require days to compute complex, high-quality impulse responses for auralization, limiting their applications.
AI training and testing was another area addressed by Treble’s technology. Its cloud-based tool generates rich, realistic datasets without traditional data collection. These simulated audio scenes support the development of systems for tasks such as speech recognition and noise suppression.
Solution:
Treble combined access to HPC with a hybrid sound simulation engine that brings together wave-based methods and geometrical acoustics. This approach supports accurate, scalable simulations across different environments. Through its cloud-based platform and software development kit, the Treble SDK, the company enables users to model sound and incorporate acoustic testing into their existing development workflows.
The same technology generates synthetic audio datasets for AI training and testing. By simulating diverse acoustic scenes, it supports audio model development for real-world applications. Virtual prototyping allows engineers to assess designs and detect acoustic problems earlier in development, helping improve equipment performance in areas such as building design and the automotive industry.
Impact and benefit:
HPC access through IHPC and DEEP helped Treble overcome limitations in the speed and scale of acoustic modelling. The reported benefits include faster simulations than traditional finite element analysis solvers, alongside improved precision. Users can evaluate designs more efficiently and identify flaws earlier, helping improve final product quality.
These capabilities reduce the need for physical prototyping and manual model preparation, with associated savings in time and cost. More efficient workflows also reduce rework and material waste in sectors such as construction and manufacturing. For AI-powered sound technologies, efficient synthetic data generation supports shorter development cycles and a faster route to market.
The benefits extend across construction, automotive engineering, consumer electronics and professional audio. By making detailed sound simulation more practical within everyday design workflows, Treble helps development teams make better-informed decisions about acoustics and product performance.
Testimonial
“Treble Technologies leverages high-performance computing (HPC) to run multiple wave-based acoustic simulations simultaneously, enabling large-scale synthetic data generation. This process allows for the rapid creation of diverse, highly realistic audio datasets, essential for training advanced machine learning models. By simulating various acoustic environments in parallel, Treble accelerates the development of AI-driven applications such as speech recognition, noise suppression, and adaptive audio technologies. The scalability and efficiency of this approach empower industries to optimize their products with unparalleled precision and speed. With the support of EDIH-IS and NCC Iceland for AI&HPC, we have been able to run our simulations on GPU-based HPC clusters. This has allowed us to perform large-scale simulations, validating our system model and technology for our latest product. We are grateful for the HPC access provided by NCC Iceland for AI&HPC, which has been instrumental in advancing our work. We are proud to have collaborated with EDIH-IS and NCC Iceland for AI&HPC, contributing to a shared success story in high-performance computing and acoustic simulation.”
Timeframe
2024
References
Treble Technologies: https://www.treble.tech/
IHPC: https://www.ihpc.is/