1. Cuda Driver Update

To use CUDA on your system, you need to have: ‣ a CUDA-capable GPU ‣ Mac OS X 10.13 ‣ the Clang compiler and toolchain installed using Xcode ‣ the NVIDIA. Mac Nvidia CUDA Updates (Top of Page are Mac Nvidia Graphics Driver Updates - later first.) CUDA 418.163 Update for macOS High Sierra (May 10, 2019) Download of Nvidia CUDA 418.163 for macOS High Sierra. The page notes 'CUDA driver update to support CUDA Toolkit 10.1 Update 1 and macOS 10.13.6'. (Here's a direct download link for CUDA 418.163.). Apple fully control drivers for Mac OS. But if Apple allows, our engineers are ready and eager to help Apple deliver great drivers for Mac OS 10.14 (Mojave). Apple’s recently released macOS 10.14 (Mojave) does not support CUDA. For CUDA developers who are on macOS 10.13, it is recommended to not upgrade to Mojave. STEP 2: Download the Driver File Download - CUDADriver-387.128-macos.dmg. STEP 3: Install Note: Quadro FX for Mac or GeForce for Mac must be installed prior to CUDA 387.128 installation. Double-click on CUDADriver-387.128-macos.dmg; Click Continue on the CUDA 9.1 Installer Welcome screen. Aug 04, 2020 To verify that your system is CUDA-capable, under the Apple menu select About This Mac, click the More Info button, and then select Graphics/Displays under the Hardware list. There you will find the vendor name and model of your graphics card.

TensorFlow GPU support requires an assortment of drivers and libraries. Tosimplify installation and avoid library conflicts, we recommend using aTensorFlow Docker image with GPU support (Linux only). This setuponly requires the NVIDIA® GPU drivers.

These install instructions are for the latest release of TensorFlow. See thetested build configurations for CUDA® and cuDNN versions touse with older TensorFlow releases.

Pip package

See the pip install guide for available packages, systems requirements,and instructions. The TensorFlow pip package includes GPU support forCUDA®-enabled cards:

This guide covers GPU support and installation steps for the latest stableTensorFlow release.

Older versions of TensorFlow

For releases 1.15 and older, CPU and GPU packages are separate:

Hardware requirements

The following GPU-enabled devices are supported:

• NVIDIA® GPU card with CUDA® architectures 3.5, 3.7, 5.2, 6.0, 6.1, 7.0 andhigher than 7.0. See the list ofCUDA®-enabledGPU cards.
• On systems with NVIDIA® Ampere GPUs (CUDA architecture 8.0) or newer,kernels are JIT-compiled from PTX and TensorFlow can take over 30 minutes tostart up. This overhead can be limited to the first start up by increasingthe default JIT cache size with: 'export CUDA_CACHE_MAXSIZE=2147483648'(seeJITCaching for details).
• For GPUs with unsupported CUDA® architectures, or to avoid JIT compilationfrom PTX, or to use different versions of the NVIDIA® libraries, see theLinux build from source guide.
• Packages do not contain PTX code except for the latest supported CUDA®architecture; therefore, TensorFlow fails to load on older GPUs whenCUDA_FORCE_PTX_JIT=1 is set. (SeeApplicationCompatibility for details.)

Software requirements

The following NVIDIA® software must be installed on your system:

• NVIDIA® GPU drivers —CUDA®10.1 requires 418.x or higher.
• CUDA® Toolkit —TensorFlow supports CUDA® 10.1 (TensorFlow >= 2.1.0)
• CUPTI ships with the CUDA®Toolkit.
• cuDNN SDK 7.6 (seecuDNN versions).
• (Optional)TensorRT 6.0 to improve latency and throughput for inference on some models.

Linux setup

The apt instructions below are the easiest way to install the required NVIDIAsoftware on Ubuntu. However, if building TensorFlow from source,manually install the software requirements listed above, and consider using a-develTensorFlow Docker image as a base.

Install CUPTI which ships withthe CUDA® Toolkit. Append its installation directory to the $LD_LIBRARY_PATHenvironmental variable:

Install CUDA with apt

This section shows how to install CUDA® 10 (TensorFlow >= 1.13.0) on Ubuntu16.04 and 18.04. These instructions may work for other Debian-based distros.

Ubuntu 18.04 (CUDA 10.1)

Ubuntu 16.04 (CUDA 10.1)

Windows setup

See the hardware requirements andsoftware requirements listed above. Read theCUDA® install guide for Windows.

Make sure the installed NVIDIA software packages match the versions listed above. Inparticular, TensorFlow will not load without the cuDNN64_7.dll file. To use adifferent version, see the Windows build from source guide.

Add the CUDA®, CUPTI, and cuDNN installation directories to the %PATH%environmental variable. For example, if the CUDA® Toolkit is installed toC:Program FilesNVIDIA GPU Computing ToolkitCUDAv10.1 and cuDNN toC:toolscuda, update your %PATH% to match:

NVIDIA® CUDA Toolkit 11.0 no longer supports development or running applications on macOS. While there are no tools which use macOS as a target environment, NVIDIA is making macOS host versions of these tools that you can launch profiling and debugging sessions on supported target platforms.

You may download all these tools here. Note that the Nsight tools provide the ability to download these macOS host versions on their respective product pages.

Please visit each tool's overview page for more information about the tool and its supported target platforms.

The macOS host tools provided are:

• Nsight Systems - a system profiler and timeline trace tool supporting Pascal and newer GPUs
• Nsight Compute - a CUDA kernel profiler supporting Volta and new GPUs
• Visual Profiler - a CUDA kernel and system profiler and timeline trace tool supporting older GPUs (see installation instructions, below)
• cuda-gdb - a GPU and CPU CUDA application debugger (see installation instructions, below)

Instructions for installing cuda-gdb on the macOS

This tar archive holds the distribution of the CUDA 11.0 cuda-gdb debugger front-end for macOS.
Native macOS debugging is not supported in this release. Remote debugging from a macOS host to other CUDA enabled targets, however, is supported.

To install:
1. Create an installation directory
   INSTALL_DIR=$HOME/cuda-gdb-darwin-11.0
   mkdir $INSTALL_DIR
   cd $INSTALL_DIR
   
2. Download the cuda-gdb-darwin-11.0.tar.gz tar archive into $INSTALL_DIR above
3. Unpack the tar archive
   tar fxvz cuda-gdb-darwin-11.0.tar.gz
4. Add the bin directory to your path
   PATH=$INSTALL_DIR/bin:$PATH
5. Run cuda-gdb --version to confirm you're picking up the correct binaries
   cuda-gdb --version

You should see the following output:

NVIDIA (R) CUDA Debugger
11.0 release
Portions Copyright (C) 2007-2020 NVIDIA Corporation
GNU gdb (GDB) 8.2
Copyright (C) 2018 Free Software Foundation, Inc.
License GPLv3+: GNU GPL version 3 or later for more information:
https://docs.nvidia.com/cuda/profiler-users-guide/index.html#visual



Notes about JRE Requirements when using Visual Profiler on the macOS

OpenJDK provides an open-source (and standards compliant) implementation of a Java compliant JVM.
Binaries are provided by various vendors such as Oracle, Azul Systems (Zulu), Amazon, Red Hat, IBM, etc.
Visual Profiler needs to use an older version of Java, specifically JRE update 151, to work correctly.
This is currently not offered by Oracle JDK but is provided by Azul Systems (Zulu).
The Bazel Build project also uses the Zulu builds of OpenJDK.

Cuda Driver Update

Download JDK 8.0.144 to get JRE update 151:
• Download version: 8u144-b01 (Zulu: 8.23.0.3) .dmg.zip.tar.gz
• Download version: Zulu 8.23.0.3 (build 1.8.0_144-b01 .zip