Groq, Inc.

Sushma Honnavara-Prasad

In one embodiment, the present disclosure includes a method of reducing power in an artificial intelligence processor. For each cycle, over a plurality of cycles, an AI model is translated into operations executable on an artificial intelligence processor. The translating is based on power parameters that correspond to power consumption and performance of the artificial intelligence processor. The AI processor is configured with the executable operations, and input activation data sets are processed. Accordingly, result sets, power consumption data, and performance data are generated and stored over the plurality of cycles. The method further includes training an AI algorithm using the stored parameters, the power consumption data, and the performance data. A trained AI algorithm outputs a plurality of optimized parameters to reduce power consumption of the AI processor. The AI model is then translated into optimized executable operations based on the plurality of optimized parameters.

Power optimization in an artificial intelligence processor

Embodiments of the present disclosure pertain to reducing power consumption in a processor circuit. In one embodiment, a processor circuit comprises a plurality of data storage modules. The plurality of data storage modules each include one or more first multibit flip flop circuits having a first power consumption per bit and one or more second flip flop circuits having a second power consumption per bit. The first multibit flip flop circuits may have more bits than the second flip flop circuits. Additionally, the first power consumption per bit may be less than the second power consumption per bit such that power consumption is reduced when the first multibit flip flop circuits are used to store bits that change with a higher frequency than bits stored in the second flip flop circuits.

Growing community of inventors

San Jose, CA, United States of America

Sushma Honnavara-Prasad