• Various security issues were addressed, for additional details on the med-high severity issues please review NVIDIA Product Security for more information
• Fixed a bug in the NVIDIA GPU Boost algorithm that could cause the Tesla P100 SXM2 GPU to become unresponsive with a “GPU has fallen of the bus” error under certain workloads. In this state, the GPU is not available for any work
• Added support for Tesla V100 GPUs
• Added support for MPS on Tesla V100 GPUs
• Added nvmlClocksThrottleReasonSwThermalSlowdown as a NVML throttle reason. This is shown as SW Thermal Slowdown in nvidia-smi -q
• Added new "Memory Max Operating Temp" to nvidia-smi and SMBPBI to report the maximum memory temperature for Tesla V100
• Added new "GPU Max Operating Temp" to nvidia-smi and SMBPBI to report the maximum GPU operating temperature for Tesla V100
• Added CUDA support to allow JIT linking of binary compatible cubins
• Fixed an issue in the driver that may cause certain applications using unified memory APIs to hang

Installing the NVIDIA 384.81 Driver and CUDA 8 on AWS G2 Instances

Note that the existing CUDA 8 installer packages contain a version of the driver (375.26) that does not support the K520 GPU and thus additional steps are required to get started with using CUDA on the AWS EC2 G2 instances. One of the easier ways to install drivers and CUDA is to use the network installation package to install the NVIDIA 384.81 driver and the toolkit. For simplicity, these instructions will refer to the steps on an Ubuntu system using the APT package manager. The same instructions will apply when using other package managers such as yum or zypper.

1. Update the CUDA network repo keys using the following command
# sudo apt-key adv --fetch-keys
http://developer.download.nvidia.com/compute/cuda/repos/ubuntu1604/x86_64/7fa2af80.pub
2. Add the CUDA network repo and update the package lists on your system to get new versions of the software and their dependencies.
# sudo sh -c 'echo "deb
http://developer.download.nvidia.com/compute/cuda/repos/ubuntu1604/x86_64 /" > /etc/apt/sources.list.d/cuda.list'
# sudo apt-get update
3. If you already have CUDA 8 installed on your instance and only need to update the NVIDIA driver, install the cuda-drivers meta-package. Then reboot the instance to complete the installation of the 384.81 NVIDIA driver.
# sudo apt-get -y --no-install-recommends install cuda-drivers
# sudo reboot
4. If you also need to install the CUDA toolkit, then install the cuda-toolkit-8-0 meta-package to download and install CUDA 8.
# sudo apt-get -y install cuda-toolkit-8-0

Refer to the Linux Installation Guide for CUDA Toolkit for more information on using runfiles or local installers to install CUDA on various Linux distributions. The guide is located at the following URL: (http://docs.nvidia.com/cuda/cuda-installation-guide-linux/index.html)

V-Series:

Tesla V100

P-Series:

Tesla P100, Tesla P40, Tesla P6, Tesla P4

K-Series:

Tesla K80, Tesla K520, Tesla K40c, Tesla K40m, Tesla K40s, Tesla K40st, Tesla K40t, Tesla K20Xm, Tesla K20m, Tesla K20s, Tesla K20c, Tesla K10, Tesla K8

C-Class:

Tesla C2075, Tesla C2070, Tesla C2050

M-Class:

M60, M40 24GB, M40, M6, M4, M2090, M2075, M2070, M2070-Q, M2050

X-Class:

Tesla X2070, Tesla X2090