Imagine your phone's graphics chip is like a car engine — it has to run fast but can't guzzle too much fuel. Right now, developers writing software for mobile and wearable devices are essentially driving blind: they have no dashboard telling them where the power is being wasted. LPGPU2 built that missing dashboard — a set of tools that show exactly how much power each part of the software uses on the GPU, plus suggestions on how to fix the wasteful bits. They even proposed a universal standard so these tools could plug into any GPU maker's hardware.
Developers building software for mobile and wearable GPUs have almost no visibility into where power is being wasted. As applications grow more complex — from AR filters to automotive displays — the gap between what the hardware can deliver and what the software demands keeps widening. Without proper power profiling tools, teams burn weeks guessing at optimizations that may not even target the real bottlenecks.
The project delivered a complete GPU power analysis toolchain: hardware power and performance counters with a prototype implementation, a standardizable API proposed to Khronos (covering OpenGL ES and OpenCL), visualization and analysis software, and multiple application use-cases demonstrating the system. In total, 12 deliverables were produced.
If you are a mobile device maker dealing with battery drain complaints from users running GPU-heavy apps — this project developed a complete performance analysis toolchain that shows exactly where power is wasted on low-power GPUs. With 80% of the consortium being industry partners, the tools were designed for real production workflows, not just academic use.
If you are an automotive supplier dealing with strict power budgets in dashboard displays and driver-assist visualizations — this project built hardware performance counters and a standardizable API that lets your engineers measure GPU power consumption at the counter level. The toolchain was validated through multiple application use-cases covering different GPU workload types.
If you are a mobile game studio dealing with thermal throttling and battery drain killing your user retention — this project created visualization tools that map power and performance bottlenecks in GPU code running on OpenGL ES and OpenCL. The API was proposed as a Khronos industry standard, meaning it could integrate with your existing graphics pipeline.
Based on available project data, the toolchain was built as part of a publicly funded Innovation Action, which typically means results are available for licensing or collaboration. The API was proposed to Khronos as an open standard, so the interface layer itself may be royalty-free. Specific licensing terms would need to be discussed with the coordinator at TU Berlin.
The project was designed with industrial scalability in mind — 4 out of 5 consortium partners are industry players, and 3 are SMEs already working in GPU tooling. The standardizable API was specifically built to interface with any GPU driver that implements it, meaning it is hardware-vendor agnostic by design.
The performance monitoring API was proposed as part of Khronos standards (OpenGL ES and OpenCL), which typically follow FRAND or royalty-free licensing. Other toolchain components and hardware counter designs likely remain with the consortium partners. IP terms should be clarified directly with TU Berlin as coordinator.
The project built a prototype hardware implementation with new power and performance counters, plus an API layer to access them. The API was designed to be standardizable so any GPU vendor could implement it in their drivers. Adoption depends on GPU manufacturers integrating the proposed counters into their chips.
This was an Innovation Action (IA), which in EU funding terms targets technology validation and demonstration. The consortium built working hardware prototypes, a complete toolchain, and validated everything through multiple application use-cases. The project ran from 2016 to 2018 and delivered 12 deliverables including the proposed Khronos standard API.
The project closed in September 2018. However, if the Khronos standard proposal was adopted, it would live on within the Khronos ecosystem. For toolchain updates or collaboration, the consortium — led by TU Berlin with 3 industry SMEs — would be the starting point for discussions.
The LPGPU2 consortium is heavily industry-driven: 4 out of 5 partners are from industry, and 3 of those are SMEs, giving an 80% industry ratio. This is unusual for EU projects and signals that the results are practically oriented. TU Berlin anchors the academic side as coordinator, while the industry partners from Germany, Greece, and the UK bring GPU tooling and embedded systems expertise. For a business looking to adopt this technology, the high industry involvement means the tools were built with commercial constraints — not just academic curiosity.
Coordinator is Technische Universitat Berlin (Germany). Use SciTransfer's matchmaking service for a warm introduction to the research team.
Want to explore licensing this GPU power analysis toolchain or collaborating with the consortium? SciTransfer can arrange a direct introduction to the right people at TU Berlin and the industry partners.
