Thermal Performance and Optical Inspection Dataset of GA102 GPUs using UV-Fluorescence and IR-Thermography Dataset

This dataset provides multimodal sensor data and optical evidence for the assessment of Thermal Interface Material (TIM) application quality in high-power electronics, specifically focusing on NVIDIA RTX 3080 Ti (GA102) GPUs. It was created to support the development of AI-enabled computer vision models for automated thermal fault triage and remote diagnostics within the framework of the Research Innovate action. The dataset includes raw high-frequency telemetry logs with over 450 sensor channels per timestamp, covering clocks, power, fan speeds, and temperatures. It also contains infrared (IR) thermography videos that monitor spatial heat flow and heatsink activation during benchmark execution for both before and after repaste conditions. Additionally, the dataset provides benchmark results and scores from Unigine Superposition 1080p Extreme across four experimental stages, documenting the load profile and performance impact. The experimental stages cover original aged TIM and new high-performance materials (KRYO33 and K5 PRO Mt. Olympos Edition) under both automatic and full fan profiles. Empirical results demonstrate significant thermal improvements, with hotspot temperature reductions of up to 18.1 degrees Celsius and VRAM junction reductions of up to 16.0 degrees Celsius under comparable workloads. This research was carried out as part of the project Data Recovery and Remote Computer Diagnostic Services with AI (Project code: ΕΚΠΑΡ01-0063510) under the framework of the Action Research – Innovate of the Operational Program Competitiveness 2021-2027, that is co-funded by the European Regional Development Fund (ERDF) and Greece.