Research Identifies Method to Potentially Double Computer Processing Speeds Using Simultaneous and Heterogeneous Multithreading

  • A method has been identified to potentially double computer processing speeds using existing hardware
  • It involves simultaneous and heterogeneous multithreading (SHMT) which uses multiple processors like CPUs, GPUs, and TPUs at the same time
  • An experiment using an ARM processor, NVIDIA GPU and Tensor Processing Unit achieved a 1.96x speedup and 51% reduction in energy consumption
  • This could reduce costs and carbon emissions from data centers by making better use of existing hardware
  • It may also reduce the need for freshwater used to cool servers
  • More research is needed to address implementation challenges and determine best applications