Nvidia claims its upcoming open-source library TensorRT-LLM will double the H100's performance for running inference on leading LLMs when it debuts next month (Dylan Martin/CRN)

Dylan Martin / CRN:

Nvidia claims its upcoming open-source library TensorRT-LLM will double the H100's performance for running inference on leading LLMs when it debuts next month  —  The AI chip giant says the open-source software library, TensorRT-LLM, will double the H100's performance for running inference …

ai chip chip h100 inference library llm llms next nvidia performance running tensorrt tensorrt-llm