Graphics workload-based power trends and optimizations using PowerArtist

This presentation is about analyzing real-world workloads on GFX RTL designs, emphasizing the discovery of optimization opportunities not easily discerned through synthetic vectors or unit-level testing. The challenge lies in efficiently processing stimuli involving tens of millions of cycles on designs of multiple million instances all within a reasonable timeframe of less than 48-hours. In PowerArtist, this can be achieved by enabling the distributed GAF (using FSDB) based reduction flow. We also utilize the average power data computed using SAIF to track relative power trends through automated regressions at the GFX IP level. This approach resulted in the discovery of significant power optimization opportunity at the GFX IP level. Furthermore, it helped us successfully identify regressions in power between two successive versions of the GFX IP using relative RTL trend tracking technique. Overall, this helps with left shifting the GFX IP level power analysis and optimization effort and avoid late RTL surprises. This directly translates into better perf-per-watt for the product.