Technical Specifications

Peak double precision floating point performance

  • 5.3 Tflops

Peak single precision floating point performance

  • 10.6 TFlops

Number of accelerators per card

  • 1


  • 3584

Memory size per board

  • 16 GB CoWoS HBM2

Memory bandwidth for board

  • 732 GB/s

Accelerator applications

  • Deep learning and HPC

Architecture features

  • Deep learning models typically take days to weeks to train, forcing scientists to make compromises between accuracy and time to deployment. The NVIDIA Tesla P4 GPU accelerator, based on the NVIDIA Pascal architecture, is designed to deliver the highest combination of single precision performance together with high memory density, as required for deep learning training.


  • Compatible with HPE SGI 8600, HPE sx40 servers

Minimum dimensions ( H x W x D)

  • 7.62 x 13.97 x 5.59 cm


  • 0.68 kg


  • For details on HPE Qualified Options Limited Warranty visit: 1-year parts, 1-year labor, and 1-year on-site support coverage. For more warranty information refer to http://h20564.www2.hpe.com/hpsc/wc/public/home
  • *
    Price may vary based on local reseller.