Abstract

This support matrix is for NVIDIA optimized frameworks. The matrix provides a single view into the supported software and specific versions that come packaged with the frameworks based on the container image.

1. 20.xx Framework Containers Support Matrix

Important: Content that is included in <<>> brackets indicates new content from the previously published version.
Note: The deep learning framework container packages follow a naming convention that is based on the year and month of the image release. For example, the 20.06 release of an image was released in June, 2020.

20.xx container images

Table 1. Software stack packaged with the 20.xx container images
  Container Image 20.06 20.03 20.02 20.01
Supported Platform Host OS DGX OS Server
  • 4.4.x, 4.3.x, and 4.1.0+ (DGX-1)1
  • 4.4.x, 4.3.x, and 4.1.0+ (DGX-2)
  • 4.99.x (DGX A100)
  • DGX EL7-20.02 for RHEL7.x
DGX OS Desktop
  • 4.4.x, 4.3.x, and 4.1.0+ (DGX Station)1
  • DGX EL7-20.02 for RHEL7.x
DGX OS Server
  • 4.4.x, 4.3.x, and 4.1.0+ (DGX-1)2
  • 4.4.x, 4.3.x, and 4.1.0+ (DGX-2)
  • DGX EL7-20.02 for RHEL7.x
DGX OS Desktop
  • 4.4.x, 4.3.x, and 4.1.0+ (DGX Station)1
  • DGX EL7-20.02 for RHEL7.x
DGX OS Server
  • 4.4.x, 4.3.x, and 4.1.0+ (DGX-1)1
  • 4.4.x, 4.3.x, and 4.1.0+ (DGX-2)
  • DGX EL7-20.02 for RHEL7.x
DGX OS Desktop
  • 4.4.x, 4.3.x, and 4.1.0+ (DGX Station)1
  • DGX EL7-20.02 for RHEL7.x
DGX OS Server
  • 4.3.x, 4.1.0+ (DGX-1)1
  • 4.3.x, 4.1.0+ (DGX-2)
DGX OS Desktop
  • 4.3.x, 4.1.0+ (DGX Station)1
NVIDIA Driver

Release 20.06 is based on CUDA <<11.0.167>>, which requires NVIDIA driver release <<450.36>>.

However, if you are running on Tesla (for example, T4 or any other Tesla board), you may use NVIDIA driver release 418.xx or 440.30.

The CUDA driver's compatibility package only supports particular drivers. 3

Release 20.03 is based on CUDA 10.2.89, which requires NVIDIA driver release 440.33.01.

However, if you are running on Tesla (for example, T4 or any other Tesla board), you may use NVIDIA driver release 396, 384.111+, 410, 418.xx or 440.30.

The CUDA driver's compatibility package only supports particular drivers. 4

Release 20.02 is based on CUDA 10.2.89, which requires NVIDIA driver release 440.33.01.

However, if you are running on Tesla (for example, T4 or any other Tesla board), you may use NVIDIA driver release 396, 384.111+, 410, 418.xx or 440.30.

The CUDA driver's compatibility package only supports particular drivers. 2

Release 20.01 is based on CUDA 10.2.89, which requires NVIDIA driver release 440.33.01.

However, if you are running on Tesla (for example, T4 or any other Tesla board), you may use NVIDIA driver release 396, 384.111+, 410, 418.xx or 440.30.

The CUDA driver's compatibility package only supports particular drivers. 2

Supported Hardware GPU Model
Base Image Container OS Ubuntu 18.04 Ubuntu 18.04 Ubuntu 18.04 Ubuntu 18.04
CUDA <<11.0.167>> 10.2.89 10.2.89 10.2.89
cuBLAS <<11.1.0.213>> 10.2.2.89 10.2.2.89 10.2.2.89
cuDNN <<8.0.1>> 7.6.5 7.6.5 7.6.5
NCCL <<2.7.5>> 2.5.6 2.5.6 2.5.6
NVIDIA Optimized Frameworks Kaldi da71f301 including 5.5 including 5.5 including
Docker image size: 6.82 GB Docker image size: 5.76 GB Docker image size: 5.53 GB Docker image size: 5.55 GB
NVCaffe   0.17.3 including 0.17.3 including 0.17.3 including
  Docker image size: 4.82 GB Docker image size: 4.82 GB Docker image size: 4.85 GB
DIGITS 6.1.1 including 6.1.1 including 6.1.1 including 6.1.1 including
Docker image size with TensorFlow: 12.4 GB
  • Docker image size with TensorFlow: 10.5 GB
  • Docker image size with Caffe: 5.06 GB
  • Docker image size with TensorFlow: 10.5 GB
  • Docker image size with Caffe: 5.06 GB
  • Docker image size with TensorFlow: 9.36 GB
  • Docker image size with Caffe: 5.01 GB
MXNet 1.6.0 including 1.6.0 including 1.6.0.rc2 including 1.5.1commit c98184806 from September 4, 2019 including
Docker image size: 9.16 GB Docker image size: 6.73 GB Docker image size: 6.59 GB Docker image size: 6.11 GB
PyTorch <<1.6.0a0+9907a3e>> including 1.5.0a0+8f84ded including 1.5.0a0+3bbb36e including 1.4.0a0+a5b4d78 including
Docker image size: 11.9 GB Docker image size: 9.41 GB Docker image size: 9.11 GB Docker image size: 9.12 GB
TensorFlow <<2.2.0>> including 1.15.2 including 2.1.0 including 1.15.2 including 2.1.0 including 1.15.2 including 2.0.0 including 1.15.0 including
Docker image size: 9.45 GB Docker image size: 11.5 GB Docker image size: 8.05 GB Docker image size: 9.53 GB Docker size image: 7.42 GB Docker size image: 9.49 GB Docker size image: 7.15 GB
  • Docker image size with Python 2.7: 7.88 GB
  • Docker image size with Python 3.6: 8.39 GB
TensorRT <<7.1.2>> including: 7.0.0 including: 7.0.0 including: 7.0.0 including:
Docker image size: 4.97 GB
  • Docker image size with Python 2.7: 4.05 GB
  • Docker image size with Python 3.6: 4.08 GB
  • Docker image size with Python 2.7: 4.04 GB
  • Docker image size with Python 3.6: 4.07 GB
  • Docker image size with Python 2.7: 4.08 GB
  • Docker image size with Python 3.6: 4.11 GB
Triton Inference Server <<1.14.0>> including <<2.0.0>> including 1.12.0 including 1.11.0 including 1.10.0 including

Docker image size for V1 API: 9.73 GB

Docker image size for V2 API: 8.68 GB

Docker image size: 6.31 GB Docker image size: 6.07 GB Docker image size: 6.16 GB
TensorFlow For Jetson TensorFlow 1.15.2 and 2.1.0 for Jetson TensorFlow 1.15.2 and 2.1.0 for Jetson TensorFlow 1.15.2 and 2.1.0 for Jetson TensorFlow 1.15.0 and 2.0.0 for Jetson

2. 19.xx Framework Containers Support Matrix

Important: Content that is included in <<>> brackets indicates new content from the previously published version.
Note: The deep learning framework container packages follow a naming convention that is based on the year and month of the image release. For example, the 19.01 release of an image was released in January, 2019.

19.xx container images

Table 2. Software stack packaged with the 19.xx container images
  Container Image 19.12 19.11 19.10 19.09 19.08 19.07 19.06 19.05 19.04 19.03 19.02 19.01
Supported Platform Host OS DGX OS Server
  • 4.1.0+ (DGX-1)5
  • 4.1.0+ (DGX-2)
DGX OS Desktop
  • 4.1.0+ (DGX Station)1
DGX OS Server
  • 4.1.0+ (DGX-1)1
  • 4.1.0+ (DGX-2)
DGX OS Desktop
  • 4.1.0+ (DGX Station)1
DGX OS Server
  • 4.1.0+ (DGX-1)1
  • 4.1.0+ (DGX-2)
DGX OS Desktop
  • 4.1.0+ (DGX Station)1
DGX OS Server
  • 4.1.0+, 4.0.4+, 3.1.2+ and 2.1.1+ (DGX-1)1
  • 4.1.0+, 4.0.1+ (DGX-2)
DGX OS Desktop
  • 4.1.0+, 4.0.4+ and 3.1.2+ (DGX Station)1
DGX OS Server
  • 4.1.0+, 4.0.4+, 3.1.2+ and 2.1.1+ (DGX-1)1
  • 4.1.0+, 4.0.1+ (DGX-2)
DGX OS Desktop
  • 4.1.0+, 4.0.4+ and 3.1.2+ (DGX Station)1
DGX OS Server
  • 4.1.0+, 4.0.4+, 3.1.2+ and 2.1.1+ (DGX-1)1
  • 4.1.0+, 4.0.1+ (DGX-2)
DGX OS Desktop
  • 4.1.0+, 4.0.4+ and 3.1.2+ (DGX Station)1
DGX OS Server
  • 4.1.0+, 4.0.4+, 3.1.2+ and 2.1.1+ (DGX-1)1
  • 4.1.0+, 4.0.1+ (DGX-2)
DGX OS Desktop
  • 4.1.0+, 4.0.4+ and 3.1.2+ (DGX Station)1
DGX OS Server
  • 4.1.0+, 4.0.4+, 3.1.2+ and 2.1.1+ (DGX-1)1
  • 4.1.0+, 4.0.1+ (DGX-2)
DGX OS Desktop
  • 4.1.0+, 4.0.4+ and 3.1.2+ (DGX Station)1
DGX OS Server
  • 4.0.4+, 3.1.2+ and 2.1.1+ (DGX-1)
  • 4.0.1+ (DGX-2)
DGX OS Desktop
  • 4.0.4+ and 3.1.2+ (DGX Station)
DGX OS Server
  • 4.0.4+, 3.1.2+ and 2.1.1+ (DGX-1)
  • 4.0.1+ (DGX-2)
DGX OS Desktop
  • 4.0.4+ and 3.1.2+ (DGX Station)
DGX OS Server
  • 4.0.4+, 3.1.2+ and 2.1.1+ (DGX-1)
  • 4.0.1+ (DGX-2)
DGX OS Desktop
  • 4.0.4+ and 3.1.2+ (DGX Station)
DGX OS Server
  • 4.0.4+, 3.1.2+ and 2.1.1+ (DGX-1)
  • 4.0.1+ (DGX-2)
DGX OS Desktop
  • 4.0.4+ and 3.1.2+ (DGX Station)
NVIDIA Driver

Release 19.12 is based on CUDA 10.2.89, which requires NVIDIA driver release 440.33.01.

However, if you are running on Tesla (for example, T4 or any other Tesla board), you may use NVIDIA driver release 396, 384.111+, 410, 418.xx or 440.30

The CUDA driver's compatibility package only supports particular drivers. 6

Release 19.11 is based on CUDA 10.2.89, which requires NVIDIA driver release 440.xx.

However, if you are running on Tesla (for example, T4 or any other Tesla board), you may use NVIDIA driver release 396, 384.111+, 410 or 418.xx.

The CUDA driver's compatibility package only supports particular drivers. 2

Release 19.10 is based on CUDA 10.1.243, which requires NVIDIA driver release 418.xx.

However, if you are running on Tesla (for example, T4 or any other Tesla board), you may use NVIDIA driver release 396, 384.111+ or 410.

The CUDA driver's compatibility package only supports particular drivers. 2

Release 19.09 is based on CUDA 10.1.243, which requires NVIDIA driver release 418.xx.

However, if you are running on Tesla (for example, T4 or any other Tesla board), you may use NVIDIA driver release 396, 384.111+ or 410.

The CUDA driver's compatibility package only supports particular drivers. 2

Release 19.08 is based on CUDA 10.1.243, which requires NVIDIA driver release 418.87.

However, if you are running on Tesla (Tesla V100, Tesla P4, Tesla P40, or Tesla P100), you may use NVIDIA driver release 384.111+ or 410.

The CUDA driver's compatibility package only supports particular drivers. 2

Release 19.07 is based on CUDA 10.1.168, which requires NVIDIA driver release 418.67.

However, if you are running on Tesla (Tesla V100, Tesla P4, Tesla P40, or Tesla P100), you may use NVIDIA driver release 384.111+ or 410.

The CUDA driver's compatibility package only supports particular drivers. 2

Release 19.06 is based on CUDA 10.1.168, which requires NVIDIA driver release 418.xx.

However, if you are running on Tesla (Tesla V100, Tesla P4, Tesla P40, or Tesla P100), you may use NVIDIA driver release 384.111+ or 410.

The CUDA driver's compatibility package only supports particular drivers. 2

Release 19.05 is based on CUDA 10.1 Update 1, which requires NVIDIA driver release 418.xx.

However, if you are running on Tesla (Tesla V100, Tesla P4, Tesla P40, or Tesla P100), you may use NVIDIA driver release 384.111+ or 410.

The CUDA driver's compatibility package only supports particular drivers. 2

Release 19.04 is based on CUDA 10.1, which requires NVIDIA driver release 418.xx.x+.

However, if you are running on Tesla (Tesla V100, Tesla P4, Tesla P40, or Tesla P100), you may use NVIDIA driver release 384.111+ or 410.

The CUDA driver's compatibility package only supports particular drivers. 2

Release 19.03 is based on CUDA 10.1, which requires NVIDIA driver release 418.xx+.

However, if you are running on Tesla (Tesla V100, Tesla P4, Tesla P40, or Tesla P100), you may use NVIDIA driver release 384.111+ or 410.

The CUDA driver's compatibility package only supports particular drivers. 2

Release 19.02 is based on CUDA 10.0, which requires NVIDIA driver release 410.72+.

However, if you are running on Tesla (Tesla V100, Tesla P4, Tesla P40, or Tesla P100), you may use NVIDIA driver release 384.111+.

The CUDA driver's compatibility package only supports particular drivers. 2

Release 19.01 is based on CUDA 10.0, which requires NVIDIA driver release 410.72+.

However, if you are running on Tesla (Tesla V100, Tesla P4, Tesla P40, or Tesla P100), you may use NVIDIA driver release 384.111+.

The CUDA driver's compatibility package only supports particular drivers. 2

Supported Hardware GPU Model
Base Image Container OS Ubuntu 18.04 Ubuntu 18.04 Ubuntu 18.04 Ubuntu 18.04 Ubuntu 18.04 Ubuntu 18.04 Ubuntu 16.04 Ubuntu 16.04 Ubuntu 16.04 Ubuntu 16.04 Ubuntu 16.04 Ubuntu 16.04
CUDA 10.2.89 10.2.89 10.1.243 10.1.243 10.1.243 10.1.168 10.1.168 10.1 Update 1 10.1.105 10.1.105 10.0.130 10.0.130
cuBLAS 10.2.2.89 10.2.2.89 10.2.1.243 10.2.1.243 10.2.1.243 10.2.0.168 10.2.0.168 10.1 Update 1 10.1.0.105 10.1.105 10.0.130 10.0.130
cuDNN 7.6.5 7.6.5 7.6.4 7.6.3 7.6.2 7.6.1 7.6.0 7.6.0 7.5.0 7.5.0 7.4.2 7.4.2
NCCL 2.5.6 2.5.6 2.4.8 2.4.8 2.4.8 2.4.7 2.4.7 2.4.6 2.4.6 2.4.3 2.3.7 2.3.7
NVIDIA Optimized Frameworks Kaldi 5.5 including 5.5 including 5.5 including 5.5 including 5.5 including 5.5 including 5.5 including 5.5 including 5.5 including 5.5 including    
Docker image size: 5.49 GB Docker image size: 5.61 GB Docker image size: 5.63 GB Docker image size: 5.57 GB Docker image size: 5.57 GB Docker image size: 5.00 GB Docker image size: 5.11 GB Docker image size: 5.11 GB Docker image size: 5.01 GB Docker image size: 5.09 GB    
NVCaffe 0.17.3 including 0.17.3 including 0.17.3 including 0.17.3 including 0.17.3 including 0.17.3 including 0.17.3 including 0.17.3 including 0.17.3 including 0.17.3 including 0.17.2 including 0.17.2 including
Docker image size: 4.81 GB Docker image size: 5.02 GB Docker image size: 5.1 GB Docker image size: 5.02 GB Docker image size: 5.02 GB Docker image size: 4.45 GB Docker image size: 4.33 GB Docker image size: 4.33 GB Docker image size: 4.29 GB Docker image size: 4.42 GB Docker image size: 3.56 GB Docker image size: 3.51 GB
DIGITS 6.1.1 including NA 6.1.1 including 6.1.1 including 6.1.1 including 6.1.1 including 6.1.1 including 6.1.1 including 6.1.1 including 6.1.1 including 6.1.1 including 6.1.1 including
  • Docker image size with TensorFlow: 9.3 GB
  • Docker image size with Caffe: 4.97 GB
NA
  • Docker image size with TensorFlow: 8.78 GB
  • Docker image size with Caffe: 5.22 GB
  • Docker image size with TensorFlow: 7.74 GB
  • Docker image size with Caffe: 5.13 GB
  • Docker image size with TensorFlow: 7.74 GB
  • Docker image size with Caffe: 5.13 GB
  • Docker image size with TensorFlow: 7.14 GB
  • Docker image size with Caffe: 4.57 GB
  • Docker image size with TensorFlow: 7.86 GB
  • Docker image size with Caffe: 4.45 GB
  • Docker image size with TensorFlow: 7.00 GB
  • Docker image size with Caffe: 4.45 GB
  • Docker image size with TensorFlow: 7.04 GB
  • Docker image size with Caffe: 4.41 GB
  • Docker image size with TensorFlow: 6.92 GB
  • Docker image size with Caffe: 4.49 GB
  • Docker image size with TensorFlow: 6.43 GB
  • Docker image size with Caffe: 3.70 GB
  • Docker image size with TensorFlow: 5.96 GB
  • Docker image size with Caffe: 3.66 GB
MXNet 1.5.1 commit c98184806 from September 4, 2019 including 1.5.1 commit c98184806 from September 4, 2019 including 1.5.1 commit c98184806 from September 4, 2019 including 1.5.0commit 006486af3 from August 28, 2019 including 1.5.0commit 75a9e187d from June 27, 2019 including 1.5.0.rc2 including upstream commits up through commit 75a9e187d from June 27, 2019 including 1.4.1 including 1.4.0 commit 87c7addcd from February 12, 2019 including 1.4.0 commit 87c7addcd from February 12, 2019 including 1.4.0 including 1.4.0.rc0 including 1.4.0.rc0 including
Docker image size: 6.05 GB Docker image size: 6.14 GB Docker image size: 6.2 GB Docker image size: 5.75 GB Docker image size: 5.75 GB Docker image size: 5.11 GB Docker image size: 4.99 GB Docker image size: 4.95 GB Docker image size: 4.9 GB Docker image size: 4.73 GB Docker image size: 3.83 GB Docker image size: 3.82 GB
PyTorch 1.4.0a0+a5b4d78 including 1.4.0a0+174e1ba including 1.3.0a0+24ae9b5 including 1.2.0 including 1.2.0a0 including upstream commits up through commit 9130ab38 from July 31, 2019 as well as a cherry-picked performance fix 9462ca29 including 1.2.0a0including upstream commits up through commit f6aac41 from June 19, 2019 including 1.1.0commit 0885dd28 from May 28, 2019 including 1.0.1commit 828a6a3b from March 31, 2019 including 1.0.1commit 9eb0f43 from March 28, 2019 including 1.1.0a0+81e025d including 1.1.0a0+c42431b including 1.0.0 including
Docker image size: 9.28 GB Docker image size: 9.21 GB Docker image size: 9.32 GB Docker image size: 9 GB Docker image size: 9 GB Docker image size: 8.33 GB Docker image size: 7.7 GB Docker image size: 7.55 GB Docker image size: 7.45 GB Docker image size: 7.71 GB Docker image size: 6.61 GB Docker image size: 7.70 GB
TensorFlow 2.0.0 including 1.15.0 including 2.0.0 including 1.15.0 including 1.14.0 including 1.14.0 including 1.14.0 including 1.14.0 including 1.13.1 including 1.13.1 including 1.13.1 including 1.13.1 including 1.13.0-rc0 including 1.12.0 including
Docker size image: 7.71 GB
  • Docker image size with Python 2.7: 7.84 GB
  • Docker image size with Python 3.6: 8.32 GB
Docker image size: 7.78
  • Docker image size with Python 2.7: 8.02 GB
  • Docker image size with Python 3.6: 8.60 GB
  • Docker image size with Python 2.7: 7.81 GB
  • Docker image size with Python 3.6: 8.38 GB
  • Docker image size with Python 2.7: 6.78 GB
  • Docker image size with Python 3.6: 7.34 GB
  • Docker image size with Python 2.7: 6.78 GB
  • Docker image size with Python 3.6: 7.34 GB
  • Docker image size with Python 2.7: 6.18 GB
  • Docker image size with Python 3.6: 6.71 GB
Docker image size: 6.88 GB Docker image size: 6.76 GB Docker image size: 6.8 GB Docker image size: 6.72 GB Docker image size: 6.06 GB Docker image size: 5.57 GB
TensorRT 6.0.1 including: 6.0.1 including: 6.0.1 including: 6.0.1 including: 5.1.5 including: 5.1.5 including: 5.1.5 including: 5.1.5 including: 5.1.2 RC including: 5.1.2 RC including: 5.0.2 including: 5.0.2 including:
  • Docker image size Python 2.7: 4.03 GB
  • Docker image size with Python 3.6: 4.06 GB
  • Docker image size Python 2.7: 3.83 GB
  • Docker image size with Python 3.6: 4.16 GB
Docker image size: 4.2 GB Docker image size: 4.4 GB Docker image size: 4.4 GB Docker image size: 3.83 GB Docker image size: 3.83 GB Docker image size: 3.83 GB Docker image size: 3.79 GB Docker image size: 3.91 GB Docker image size: 3.01 GB Docker image size: 3.00 GB
TensorRT Inference Server 1.9.0 including 1.8.0 including 1.7.0 including 1.6.0 including 1.5.0 including 1.4.0 including 1.3.0 including 1.2.0 including 1.1.0 including 1.0.0 including 0.11.0 Beta including 0.10.0 Beta including
Docker image size: 6.12 GB Docker image size: 6.15 GB Docker image size: 8.26 GB Docker image size: 7.73 GB Docker image size: 7.73 GB Docker image size: 7.15 GB Docker image size: 7.15 GB Docker image size: 7.02 GB Docker image size: 5.22 GB Docker image size: 5.33 GB Docker image size: 4.42 GB Docker image size: 4.17 GB
TensorFlow For Jetson TensorFlow 1.15.0 and 2.0.0 for Jetson TensorFlow 1.15.0 and 2.0.0 for Jetson TensorFlow 1.14.0 for Jetson TensorFlow 1.14.0 for Jetson TensorFlow 1.14.0 for Jetson TensorFlow 1.14.0 for Jetson   TensorFlow 1.13.1 for Jetson TensorFlow 1.13.1 for Jetson TensorFlow 1.13.1 for Jetson TensorFlow 1.13.0-rc0 for Jetson TensorFlow 1.12.0 for Jetson

18.xx Framework Containers Support Matrix

Note: The deep learning framework container packages follow a naming convention that is based on the year and month of the image release. For example, the 18.01 release of an image was released in January, 2018.

18.xx container images

Table 3. Software stack packaged with the 18.xx container images
  Container Image 18.12 18.11 18.10 18.09 18.08 18.07 18.06 18.05 18.04 18.03 18.02 18.01
Supported Platform Host OS DGX OS Server
  • 4.0.4+, 3.1.2+ and 2.1.1+ (DGX-1)
  • 4.0.1+ (DGX-2)
DGX OS Desktop
  • 4.0.4+ and 3.1.2+ (DGX Station)
DGX OS Server
  • 4.0.4+, 3.1.2+ and 2.1.1+ (DGX-1)
  • 4.0.1+ (DGX-2)
DGX OS Desktop
  • 4.0.4+ and 3.1.2+ (DGX Station)
DGX Software Stack for Red Hat Enterprise Linux
  • EL7-18.11 (DGX-1)
DGX OS Server
  • 3.1.2+ and 2.1.1+ (DGX-1)
  • 4.0.1+ (DGX-2)
DGX OS Desktop
  • 3.1.2+ (DGX Station)
DGX OS Server
  • 3.1.2+ and 2.1.1+ (DGX-1)
  • 4.0.1+ (DGX-2)
DGX OS Desktop
  • 3.1.2+ (DGX Station)
DGX OS Server
  • 3.1.2+ and 2.1.1+ (DGX-1)
  • 4.0.1+ (DGX-2)
DGX OS Desktop
  • 3.1.2+ (DGX Station)
DGX OS Server
  • 3.1.2+ and 2.1.1+ (DGX-1)
  • 4.0.1+ (DGX-2)
DGX OS Desktop
  • 3.1.2+ (DGX Station)
DGX OS Server
  • 3.1.2+ and 2.1.1+ (DGX-1)
  • 4.0.1+ (DGX-2)
DGX OS Desktop
  • 3.1.2+ (DGX Station)
DGX OS Server
  • 3.1.2+ and 2.1.1+ (DGX-1)
  • 4.0.1+ (DGX-2)
DGX OS Desktop
  • 3.1.2+ (DGX Station)
DGX OS Server
  • 3.1.2+ and 2.1.1+ (DGX-1)
  • 4.0.1+ (DGX-2)
DGX OS Desktop
  • 3.1.2+ (DGX Station)
DGX OS Server
  • 3.1.2+ and 2.1.1+ (DGX-1)
  • 4.0.1+ (DGX-2)
DGX OS Desktop
  • 3.1.2+ (DGX Station)
DGX OS Server
  • 3.1.2+ and 2.1.1+ (DGX-1)
  • 4.0.1+ (DGX-2)
DGX OS Desktop
  • 3.1.2+ (DGX Station)
DGX OS Server
  • 3.1.2+ and 2.1.1+ (DGX-1)
  • 4.0.1+ (DGX-2)
DGX OS Desktop
  • 3.1.2+ (DGX Station)
NVIDIA Driver

Release 18.12 is based on CUDA 10.0, which requires NVIDIA driver release 410.72+.

However, if you are running on Tesla (Tesla V100, Tesla P4, Tesla P40, or Tesla P100), you may use NVIDIA driver release 384.111+.

The CUDA driver's compatibility package only supports particular drivers (see footnote 1).

Release 18.11 is based on CUDA 10.0, which requires NVIDIA driver release 410.72+.

However, if you are running on Tesla (Tesla V100, Tesla P4, Tesla P40, or Tesla P100), you may use NVIDIA driver release 384.111+.

The CUDA driver's compatibility package only supports particular drivers (see footnote 1).

Release 18.10 is based on CUDA 10.0, which requires NVIDIA driver release 410.72+.

However, if you are running on Tesla (Tesla V100, Tesla P4, Tesla P40, or Tesla P100), you may use NVIDIA driver release 384.111+.

The CUDA driver's compatibility package only supports particular drivers (see footnote 1).

Release 18.09 is based on CUDA 10.0, which requires NVIDIA driver release 410.72+.

However, if you are running on Tesla (Tesla V100, Tesla P4, Tesla P40, or Tesla P100), you may use NVIDIA driver release 384.111+.

The CUDA driver's compatibility package only supports particular drivers (see footnote 1).

Ubuntu 16.04
NVIDIA Driver 384.xx+
Ubuntu 16.04
NVIDIA Driver 384.xx+
Ubuntu 16.04
NVIDIA Driver 384.xx+
Ubuntu 16.04
NVIDIA Driver 384.xx+
Ubuntu 16.04
NVIDIA Driver 384.xx+
Ubuntu 16.04
NVIDIA Driver 384.xx+
Ubuntu 16.04
NVIDIA Driver 384.xx+
Ubuntu 16.04
NVIDIA Driver 384.xx+
Supported Hardware GPU Model Volta and Pascal Volta and Pascal Volta and Pascal Volta and Pascal Volta and Pascal Volta and Pascal Volta and Pascal Volta and Pascal
Base Image Container OS Ubuntu 16.04 Ubuntu 16.04 Ubuntu 16.04 Ubuntu 16.04 Ubuntu 16.04 Ubuntu 16.04 Ubuntu 16.04 Ubuntu 16.04 Ubuntu 16.04 Ubuntu 16.04 Ubuntu 16.04 Ubuntu 16.04
CUDA 10.0.130 10.0.130 10.0.130 10.0.130 includes:
  • Support for DGX-2
  • Support for Turing
  • Support for Jetson Xavier
  • CUDA 10 compatibility package version 410.487
9.0.176 9.0.176 9.0.176 9.0.176 9.0.176 9.0.176 9.0.176 9.0.176
cuBLAS 10.0.130 10.0.130 10.0.130 10.0.130 9.0.425 9.0.425 9.0.333 9.0.333 9.0.333 9.0.333 9.0.282 Patch 2 and cuBLAS 9.0.234 Patch 1 9.0.282 Patch 2
cuDNN 7.4.1 7.4.1 7.4.0 7.3.0 7.2.1 7.1.4 7.1.4 7.1.2 7.1.1 7.1.1 7.0.5 7.0.5
NCCL 2.3.7 2.3.7 2.3.6 2.3.4 2.2.13 2.2.13 2.2.13 2.1.15 2.1.15 2.1.2 2.1.2 2.1.2
NVIDIA Optimized Frameworks NVCaffe 0.17.2 including 0.17.1 including 0.17.1 including 0.17.1 including 0.17.1 including 0.17.1 and Python 2.7 0.17.0 and Python 2.7 0.17.0 and Python 2.7 0.17.0 and Python 2.7 0.16.6 and Python 2.7 0.16.5 and Python 2.7 0.16.5 and Python 2.7
Docker image size: 3.41 GB Docker image size: 3.41 GB Docker image size: 3.41 GB Docker image size: 3.40 GB Docker image size: 3.37 GB Docker image size: 4.29 GB
Caffe2         0.8.1 including 0.8.1 including 0.8.1 including 0.8.1 including 0.8.1 including 0.8.1 including 0.8.1 including 0.8.1 including
        Docker image size: 3.02 GB Docker image size: 2.94 GB
DIGITS 6.1.1 including 6.1.1 including 6.1.1 including 6.1.1 including 6.1.1 including 6.1.1 including 6.1.1 including 6.1.1 including 6.1.1 including 6.1.0 including 6.1.0 including 6.0.0 including
  • TensorFlow Docker image size: 5.03 GB
  • Caffe Docker image size: 3.56 GB
  • TensorFlow Docker image size: 5.03 GB
  • Caffe Docker image size: 3.56 GB
Docker image size: 6.17 GB Docker image size: 5.33 GB Docker image size: 6.20 GB Docker image size: 7.16 GB
Microsoft Cognitive Toolkit         2.5 including 2.5 including 2.5 including 2.5 including 2.4 including 2.4 including 2.3.1 including 2.3.1 including
        Docker image size: 6.17 GB Docker image size: 6.13 GB
MXNet 1.3.1 including 1.3.0 including 1.3.0 including 1.3.0 including 1.2.0 including 1.2.0 including 1.2.0 including 1.1.0 including 1.1.0 including 1.1.0 including 1.0.0 including 1.0.0 including
Docker image size: 3.69 GB Docker image size: 3.69 GB Docker image size: 3.68 GB Docker image size: 3.58 GB Docker image size: 4.09 GB Docker image size: 3.93 GB
PyTorch 0.4.1+ including 0.4.1+ including 0.4.1+ including 0.4.1+ including 0.4.1 including 0.4.0 including 0.4.0 including 0.4.0 including 0.3.1 and Python 3.6 0.3.0 and Python 3.6 0.3.0 and Python 3.6 0.3.0 and Python 3.6
Docker image size: 6.08 GB Docker image size: 6.08 GB Docker image size: 6.00 GB Docker image size: 5.89 GB Docker image size: 5.64 GB Docker image size: 5.67 GB
TensorFlow 1.12.0 including 1.12.0-rc2 including 1.10.0 including 1.10.0 including 1.9.0 including 1.8.0 including 1.8.0 including 1.7.0 including 1.7.0 including 1.4.0 including 1.4.0 including 1.4.0 including
Docker image size: 4.64 GB Docker image size: 4.64 GB Docker image size: 4.57 GB Docker image size: 3.75 GB Docker image size: 3.40 GB Docker image size: 3.34 GB
TensorFlow For Jetson TensorFlow 1.12.0 for Jetson TensorFlow 1.12.0-rc2 for Jetson                    
TensorRT TensorRT 5.0.2 including: TensorRT 5.0.2 including: TensorRT 5.0.0 RC including: TensorRT 5.0.0 RC including Python 2.7 or Python 3.5 4.0.1 and Python 2.7 or Python 3.5 4.0.1 and Python 2.7 or Python 3.5 4.0.1 and Python 2.7 or Python 3.5 3.0.4 and Python 2.7 3.0.4 and Python 2.7 3.0.4 and Python 2.7 3.0.4 and Python 2.7 3.0.1 and Python 2.7
Docker image size: 3.00 GB Docker image size: 3.00 GB Docker image size: 2.99 GB Docker image size: 2.98 GB Docker image size: 2.56 GB Docker image size: 2.61 GB
TensorRT Inference Server 0.9.0 Beta including 0.8.0 Beta including 0.7.0 Beta including 0.6.0 Beta including 0.5.0 Beta including 0.4.0 Beta including 0.3.0 Beta including 0.2.0 Beta including 0.1.0 Beta      
Docker image size: 4.17 GB Docker image size: 4.17 GB Docker image size: 4.15 GB Docker image size: 4.42 GB Docker image size: 2.37 GB Docker image size: 2.47 GB
Theano         1.0.2 and Python 2.7 1.0.2 and Python 2.7 1.0.1 and Python 2.7 1.0.1 and Python 2.7 1.0.1 and Python 2.7 1.0.1 and Python 2.7 1.0.1 and Python 2.7 1.0.1 and Python 2.7
        Docker image size: 3.70 GB Docker image size: 3.74 GB
Torch         7 and Python 2.7 7 and Python 2.7 7 and Python 2.7 7 and Python 2.7 7 and Python 2.7 7 and Python 2.7 7 and Python 2.7 7 and Python 2.7
        Docker image size: 3.06 GB Docker image size: 3 GB

17.xx Framework Containers Support Matrix

Note: The deep learning framework container packages follow a naming convention that is based on the year and month of the image release. For example, the 17.01 release of an image was released in January, 2017.

17.xx container images

Table 4. Software stack packaged with the 17.xx container images
  Container Image 17.12 17.11 17.10 17.09 17.07 17.06 17.05 17.04 17.03 17.02 17.01
Supported Platform DGX OS 3.1.2+ and 2.1.1+ 3.1.2+ and 2.1.1+ 3.1.2+ and 2.1.1+ 3.1.2+ and 2.1.1+ 2.x+ and 1.x+ 2.x+ and 1.x+ 2.x+ and 1.x+ 2.x+ and 1.x+ 2.x+ and 1.x+ 2.x+ and 1.x+ 2.x+ and 1.x+
NVIDIA Driver 384 384 384 384              
Base Image Ubuntu 16.04 16.04 16.04 16.04 16.04 16.04 16.04 16.04 16.04 14.04 14.04
CUDA 9.0.176 9.0.176 9.0 9.0 8.0.61.2 8.0.61 8.0.61 8.0.61 8.0.61 8.0.61 8.0.54
cuBLAS 9.0.234 9.0.234     patch 2            
cuDNN 7.0.5 7.0.4 7.0.3 7.0.2 6.0.21 6.0.21 6.0.21 6.0.20 6.0.20 6.0.13 6.0.10
NCCL 2.1.2 2.1.2 2.0.5 2.0.5 2.0.3 1.6.1 1.6.1 1.6.1 1.6.1 1.6.1 1.6.1
NVIDIA Optimized Frameworks NVCaffe 0.16.4 0.16.4 0.16.4 0.16.4 0.16 0.16 0.16 0.16 0.16 0.16 0.16
Caffe2 0.8.1 and OpenMPI 1.10.3 0.8.1 and OpenMPI 1.10.3 0.8.1 and OpenMPI 1.10.3 0.8.1 and OpenMPI 1.10.3 0.7.0 and OpenMPI 1.10.3 0.7.0 and OpenMPI 1.10.3 0.5.0+ and OpenMPI 1.10.3 0.5.0+ and OpenMPI 1.10.3      
DIGITS 6.0.0 including 6.0.0 including 6.0.0 including 6.0.0 including 6.0.0 including 5.0 including 5.0 including 5.0 including 5.0 including 5.0 including 5.0 including
Microsoft Cognitive Toolkit 2.2 and OpenMPI 3.0.0 2.2 and OpenMPI 3.0.0 2.2 2.1 2.0 2.0 2.0.rc2 2.0.beta15.0 2.0.beta12.0 2.0.beta9.0 2.0.beta5.0
MXNet 1.0.0 0.12.0 0.11.0 0.11.0.rc3 0.10.0 0.10.0 0.9.3a+ 0.9.3a+ 0.9.3    
PyTorch 0.2.0 0.2.0 0.2.0 0.2.0 0.1.12 0.1.12 0.1.12 0.1.10      
TensorFlow 1.4.0 1.3.0 1.3.0 1.3.0 1.2.1 1.1.0 1.0.1 1.0.1 1.0.0 0.12.1 0.12.0
TensorRT 3.0.1                    
Theano 1.0.0rc1 1.0.0rc1 0.10beta3 0.10beta1 0.9.0 0.9.0 0.9.0 0.9.0 0.9.0rc3 0.8.0 0.8.0
Torch 7 7 7 7 7 7 7 7 7 7 7

16.xx Framework Containers Support Matrix

Note: The deep learning framework container packages follow a naming convention that is based on the year and month of the image release. For example, the 16.12 release of an image was released in December, 2016.

16.xx container images

Table 5. Software stack packaged with the 16.xx container images
  Container Image 16.12
Supported Platform DGX OS 2.x+ and 1.x+
NVIDIA Driver  
Base Image Ubuntu 14.04
CUDA 8.0.54
cuBLAS  
cuDNN 6.0.5
NCCL 1.6.1
NVIDIA Optimized Frameworks NVCaffe 0.16
Caffe2  
DIGITS 5.0 including
Microsoft Cognitive Toolkit 2.0.beta5.0
MXNet  
PyTorch  
TensorFlow 0.12.0
TensorRT  
Theano 0.8.0
Torch 7

Notices

Notice

This document is provided for information purposes only and shall not be regarded as a warranty of a certain functionality, condition, or quality of a product. NVIDIA Corporation (“NVIDIA”) makes no representations or warranties, expressed or implied, as to the accuracy or completeness of the information contained in this document and assumes no responsibility for any errors contained herein. NVIDIA shall have no liability for the consequences or use of such information or for any infringement of patents or other rights of third parties that may result from its use. This document is not a commitment to develop, release, or deliver any Material (defined below), code, or functionality.

NVIDIA reserves the right to make corrections, modifications, enhancements, improvements, and any other changes to this document, at any time without notice.

Customer should obtain the latest relevant information before placing orders and should verify that such information is current and complete.

NVIDIA products are sold subject to the NVIDIA standard terms and conditions of sale supplied at the time of order acknowledgement, unless otherwise agreed in an individual sales agreement signed by authorized representatives of NVIDIA and customer (“Terms of Sale”). NVIDIA hereby expressly objects to applying any customer general terms and conditions with regards to the purchase of the NVIDIA product referenced in this document. No contractual obligations are formed either directly or indirectly by this document.

NVIDIA products are not designed, authorized, or warranted to be suitable for use in medical, military, aircraft, space, or life support equipment, nor in applications where failure or malfunction of the NVIDIA product can reasonably be expected to result in personal injury, death, or property or environmental damage. NVIDIA accepts no liability for inclusion and/or use of NVIDIA products in such equipment or applications and therefore such inclusion and/or use is at customer’s own risk.

NVIDIA makes no representation or warranty that products based on this document will be suitable for any specified use. Testing of all parameters of each product is not necessarily performed by NVIDIA. It is customer’s sole responsibility to evaluate and determine the applicability of any information contained in this document, ensure the product is suitable and fit for the application planned by customer, and perform the necessary testing for the application in order to avoid a default of the application or the product. Weaknesses in customer’s product designs may affect the quality and reliability of the NVIDIA product and may result in additional or different conditions and/or requirements beyond those contained in this document. NVIDIA accepts no liability related to any default, damage, costs, or problem which may be based on or attributable to: (i) the use of the NVIDIA product in any manner that is contrary to this document or (ii) customer product designs.

No license, either expressed or implied, is granted under any NVIDIA patent right, copyright, or other NVIDIA intellectual property right under this document. Information published by NVIDIA regarding third-party products or services does not constitute a license from NVIDIA to use such products or services or a warranty or endorsement thereof. Use of such information may require a license from a third party under the patents or other intellectual property rights of the third party, or a license from NVIDIA under the patents or other intellectual property rights of NVIDIA.

Reproduction of information in this document is permissible only if approved in advance by NVIDIA in writing, reproduced without alteration and in full compliance with all applicable export laws and regulations, and accompanied by all associated conditions, limitations, and notices.

THIS DOCUMENT AND ALL NVIDIA DESIGN SPECIFICATIONS, REFERENCE BOARDS, FILES, DRAWINGS, DIAGNOSTICS, LISTS, AND OTHER DOCUMENTS (TOGETHER AND SEPARATELY, “MATERIALS”) ARE BEING PROVIDED “AS IS.” NVIDIA MAKES NO WARRANTIES, EXPRESSED, IMPLIED, STATUTORY, OR OTHERWISE WITH RESPECT TO THE MATERIALS, AND EXPRESSLY DISCLAIMS ALL IMPLIED WARRANTIES OF NONINFRINGEMENT, MERCHANTABILITY, AND FITNESS FOR A PARTICULAR PURPOSE. TO THE EXTENT NOT PROHIBITED BY LAW, IN NO EVENT WILL NVIDIA BE LIABLE FOR ANY DAMAGES, INCLUDING WITHOUT LIMITATION ANY DIRECT, INDIRECT, SPECIAL, INCIDENTAL, PUNITIVE, OR CONSEQUENTIAL DAMAGES, HOWEVER CAUSED AND REGARDLESS OF THE THEORY OF LIABILITY, ARISING OUT OF ANY USE OF THIS DOCUMENT, EVEN IF NVIDIA HAS BEEN ADVISED OF THE POSSIBILITY OF SUCH DAMAGES. Notwithstanding any damages that customer might incur for any reason whatsoever, NVIDIA’s aggregate and cumulative liability towards customer for the products described herein shall be limited in accordance with the Terms of Sale for the product.

VESA DisplayPort

DisplayPort and DisplayPort Compliance Logo, DisplayPort Compliance Logo for Dual-mode Sources, and DisplayPort Compliance Logo for Active Cables are trademarks owned by the Video Electronics Standards Association in the United States and other countries.

HDMI

HDMI, the HDMI logo, and High-Definition Multimedia Interface are trademarks or registered trademarks of HDMI Licensing LLC.

OpenCL

OpenCL is a trademark of Apple Inc. used under license to the Khronos Group Inc.

Trademarks

NVIDIA, the NVIDIA logo, and cuBLAS, CUDA, cuDNN, DALI, DIGITS, DGX, DGX-1, DGX-2, DGX Station, DLProf, Jetson, Kepler, Maxwell, NCCL, Nsight Compute, Nsight Systems, NvCaffe, NVIDIA Ampere GPU Architecture, PerfWorks, Pascal, SDK Manager, Tegra, TensorRT, Triton Inference Server, Tesla, TF-TRT, and Volta are trademarks and/or registered trademarks of NVIDIA Corporation in the U.S. and other countries. Other company and product names may be trademarks of the respective companies with which they are associated.

1 Ensure you upgrade your DGX systems to DGX OS Server version 4.1 or greater or DGX OS Desktop version 4.1 or greater to obtain security and feature updates.
2 Ensure you upgrade your DGX systems to DGX OS Server version 4.1 or greater or DGX OS Desktop version 4.1 or greater to obtain security and feature updates.
3 For a complete list of supported drivers, see the CUDA Application Compatibility topic. For more information, see CUDA Compatibility and Upgrades.
4 For a complete list of supported drivers, see the CUDA Application Compatibility topic. For more information, see CUDA Compatibility and Upgrades.
5 We will stop supporting DGX OS 3.x in September 2019. Ensure you upgrade your DGX systems to DGX OS Server version 4.1 or greater or DGX OS Desktop version 4.1 or greater to obtain security and feature updates.
6 For a complete list of supported drivers, see the CUDA Application Compatibility topic. For more information, see CUDA Compatibility and Upgrades.
7 The compatibility package ensures that Linux drivers R384 are compatible with Tesla GPUs.