Worked on enhancing hardware benchmarking capabilities in the Lightning-AI/pytorch-lightning repository by expanding the performance metrics reporting pipeline to support the RTX 4080 Super GPU. Focused on integrating accurate CUDA FLOPS values for multiple data types, including float32, tfloat32, bfloat16, float16, int8, and int4, into the existing _CUDA_FLOPS dictionary. Used Python to ensure seamless and efficient integration with current instrumentation, enabling more precise performance optimization and traceability. The work addressed the need for up-to-date hardware support in benchmarking tools, contributing a targeted feature that improves the reliability of performance metrics for modern GPU architectures.