HPE Media Server can use a graphics card (GPU) to perform some processing tasks. Using a GPU rather than the CPU can significantly increase the speed of training and analysis tasks that use Convolutional Neural Networks.
Tasks that use Convolutional Neural Networks are:
Before starting, ensure that your server has a GPU that meets the system requirements.
HPE Media Server with GPU support is available only on the Linux x86-64 platform and is supported only with Ubuntu 16.04 and Ubuntu 14.04. You can download HPE Media Server with GPU support from the Big Data Download Center. The HPE Media Server executable installed by the IDOL Server installer does not include GPU support.
HPE recommends that you install a headless version of Ubuntu 14.04.
To install HPE Media Server, run the installation program and follow the on-screen instructions.
To use a GPU to accelerate processing, you must install the NVIDIA CUDA driver version 7.5, which is included in the CUDA Toolkit version 7.5.
The following procedure explains how to install the NVIDIA CUDA driver. For more detailed information about how to install the driver, refer to the NVIDIA CUDA driver installation instructions.
To install the NVIDIA CUDA driver
Verify that your machine is running a supported operating system. Run the following command:
lsb_release -a
The operating system is described:
Distributor ID: Ubuntu Description: Ubuntu 14.04.3 LTS Release: 14.04 Codename: trusty
Verify that a CUDA-compatible card is available. Run the following command:
lspci | grep -i nvidia | grep -i VGA
This should produce output similar to:
0f:00.0 VGA compatible controller: NVIDIA Corporation GF106GL [Quadro 2000] (rev a1) 28:00.0 VGA compatible controller: NVIDIA Corporation GK110GL [Quadro K6000] (rev a1)
You must verify that one or more of these GPUs support CUDA version 3.0 to 5.0. In the previous example, the Quadro 2000 GPU is not supported.
Install the essential utilities required to install the NVIDIA driver, by running the following command. For more information, refer to the NVIDIA CUDA driver installation instructions.
sudo apt-get install build-essential
Install the NVIDIA driver by running the following commands.
sudo apt-get purge nvidia* sudo add-apt-repository ppa:graphics-drivers/ppa sudo apt-get update sudo apt-get install nvidia-352 sudo reboot
Verify that the driver installation was successful by running the nvidia-smi
command. The driver version must be reported as the following version (or later):
352.07
To determine whether HPE Media Server can use your GPU, start HPE Media Server and open the application log. The log should show the following after HPE Media Server starts:
GPU devices available for use GPU deviceID 0 (Hex PCIBusId: ..) GPU deviceID 1 (Hex PCIBusId: ..) GPU deviceID 2 (Hex PCIBusId: ..) GPU deviceID 3 (Hex PCIBusId: ..) GPU mode used for MediaServer GPU Memory: ... bytes free ... bytes total ... bytes free to allocate GPU deviceID currently used: 2 (Hex PCIBusId: ..)
If the machine has more than one GPU, you might need to set the configuration parameter GPUDeviceID
, to specify which GPU to use. To find out which GPU device ID corresponds to which GPU, match the PCI Bus ID logged by HPE Media Server to the PCI Bus ID output by the nvidia-smi
command, which also provides the GPU name.
If installation was unsuccessful, the application log shows this message instead:
CPU mode used for MediaServer
|