Automatic Generation of Microprocessor Power/Thermal Stress Benchmarks

Background Microprocessor designers are interested in estimating the maximum power and maximum temperature that the chip can reach, as it helps in designing the power delivery system, packaging, cooling, and power/thermal management schemes. However, due to the lack of any standardized stress benchmarks, the current trend in the industry is to develop hand-coded synthetic "max power" benchmarks or stress marks. This is an extremely tedious and time-consuming task that requires an intimate understanding of the processor.

Invention Description Researchers at The University of Texas at Austin have developed an automated method that utilizes machine learning principles to generate stress benchmarks that will exercise the microprocessor to its maximum power and thermal characteristics. The proposed Stressmaker framework will enable the automatic exploration of the large temperature and design space, thus addressing the problems that occur in developing hand-coded benchmarks.

Benefits

Does not require a skilled designer to create benchmarks Reduces the time required to design the program for creation of benchmarks Enables the microprocessor to exercise its maximum power/temperature Enables generating stress marks for cases where manually writing a stress mark is difficult because of the complex hardware-software interactions in today's high-performance microprocessors.

Market Potential/Applications Microprocessor industry

Development Stage Proof of concept

IP Status One U.S. patent application filed

UT Researcher Lizy K. John, Ph.D., Electrical and Computer Engineering, The University of Texas at Austin Ajay M. Joshi, The University of Texas at Austin

Type of Offer: Licensing



« More Computer Science Patents

---