Comparable SCI Benchmark Proposal #359
Description
In the past the WG has discussed the challenges with calculating a SCI score that can be compared to other scores. The WG is discussing an approach where standardized hardware and workloads are defined that software can be tested on to make different software with the same core functionality comparable. We are scoping out what it would take to establish a standard like this.
The WG is actively working taking notes here:
https://docs.google.com/document/d/1rCNbwKiegUorrtuBuw-Ywdum1rHBWfAL-gsaOmGCcpc/edit#heading=h.3skgaml8cvuo
This addresses questions raised in:
#340 Define Letter (A to G) Based Energy Ratings for a Given Software Application
#233 Investigate a software procurement standard
#326 Repeatability of Energy measurements
#208 Software system boundary requirements for comparing scores
#308 Design Time versus Runtime - I as a constant
#348 Research project on Sustainable Computing
#260 Could we calculate the impact of Algorithm selection on overall carbon emission of the software?
#253 Using SCI to measure purchased software
#244 SCI depends on hardware and usage
#189 Consider how uncertainty from simplification will impact the goals of transferability and comparability