Abstract

These support matrices provide a look into the supported platforms, features, and hardware capabilities of the TensorRT 8.0.3 APIs, parsers, and layers.

For previously released TensorRT documentation, see TensorRT Archives.

1. Features For Platforms And Software

This section lists the supported TensorRT features based on which platform and software.
Table 1. List of supported features per platform.
  Linux x86-64 Windows x64 Linux ppc64le Linux AArch64
8.0.0 EA 8.0.x GA 8.0.x 8.0.x 8.0.x
Supported CUDA versions

11.31

11.0 update 1

10.2

11.3 update 12

11.2 update 22

11.1 update 12

11.0 update 12

10.2

11.3 update 1

11.2 update 2

11.1 update 1

11.0 update 1

10.2

11.3 update 1

11.3 update 1

10.2

Supported cuBLAS versions

11.4.2.10064

11.2.0.252

10.2.3.254

11.5.1.109

11.4.1.1043

11.3.0.106

11.2.0.252

10.2.3.254

11.5.1.109

11.4.1.1043

11.3.0.106

11.2.0.252

10.2.3.254

11.5.1.109

11.5.1.109

10.2.2.214

Supported cuDNN versions cuDNN 8.2.0 cuDNN 8.2.1 cuDNN 8.2.1 cuDNN 8.2.1 cuDNN 8.2.1
TensorRT Python API Yes Yes No Yes Yes
NvUffParser Yes Yes Yes Yes Yes
NvOnnxParser Yes Yes Yes Yes Yes
Loops Yes Yes Yes Yes Yes
Note:
  • Serialized engines are not portable across platforms or TensorRT versions.
  • Refer to the minimum compatible driver versions in the CUDA Release Notes for specific NVIDIA Driver versions.

2. Layers And Features

The section lists the supported TensorRT layers and each of the features.

About this task

Note:
  • Supports broadcast indicates support for broadcast in this layer. This layer allows its two input tensors to be of dimensions [1, 5, 4, 3] and [1, 5, 1, 1], and its output is [1, 5, 4, 3]. The second input tensor has been broadcast in the innermost 2 dimensions.
  • Supports broadcast across batch indicates support for broadcast across the batch dimension. “NA” in this column means it's not allowed in networks with an implicit batch dimension.
Table 2. List of supported features per TensorRT layer.
Layer Dimensions of input tensor Dimensions of output tensor Does the operation apply to only the innermost 3 dimensions? Supports broadcast Supports broadcast across batch
IActivationLayer 0-7 dimensions 0-7 dimensions No No No
IConcatenationLayer 1-7 dimensions 1-7 dimensions No No No
IConstantLayer has no inputs 0-7 dimensions No No Always
IConvolutionLayer > 2D Convolution 3 or more dimensions 3 or more dimensions Yes No No
IConvolutionLayer > 3D Convolution 4 or more dimensions 4 or more dimensions No No No
IDeconvolutionLayer > 2D Deconvolution 3 or more dimensions 3 or more dimensions Yes No No
IDeconvolutionLayer > 3D Deconvolution 4 or more dimensions 4 or more dimensions No No No
IDequantizeLayer 2 or more dimensions 2 or more dimensions Yes No No
IElementWiseLayer 0-7 dimensions 0-7 dimensions No Yes Yes
IFillLayer 1 dimension 0-7 dimensions No NA NA
IFullyConnectedLayer 3 or more dimensions 3 or more dimensions Yes No No
IGatherLayer
  • Input1: 1-7 dimensions
  • Input2: 0-7 dimensions
0-7 dimensions No No Yes
IIdentityLayer 0-7 dimensions 0-7 dimensions No No No
IIteratorLayer 1-7 dimensions 0-6 dimensions No No NA
ILoopOutputLayer 0-7 dimensions 0-7 dimensions No No NA
ILRNLayer 3 or more dimensions 3 or more dimensions Yes No No
IMatrixMultiplyLayer 2 or more dimensions 2 or more dimensions No Yes Yes
IPaddingLayer 3 or more dimensions 3 or more dimensions Yes No No
IParametricReluLayer 1-7 dimensions 1-7 dimensions No No No
IPluginV2Layer User defined User defined User defined User defined User defined
IPoolingLayer > 2D Pooling 3 or more dimensions 3 or more dimensions Yes Yes Yes
IPoolingLayer > 3D Pooling 4 or more dimensions 4 or more dimensions No Yes Yes
IQuantizeLayer 2 or more dimensions 2 or more dimensions Yes No No
IRaggedSoftMaxLayer
  • Input: 2 dimensions
  • Bounds: 2 dimensions
2 or more dimensions No No Yes
IRecurrenceLayer 0-7 dimensions 0-7 dimensions No No NA
IReduceLayer 1-7 dimensions 0-7 dimensions No No No
IResizeLayer 1-7 dimensions 1-7 dimensions No No No
IRNNv2Layer
  • Data/Hidden/Cell: 2 or more dimensions
  • Seqlen: 0 or more dimensions
Data/Hidden/Cell: 2 or more dimensions No No No
IScaleLayer 3 or more dimensions 3 or more dimensions Yes No No
ISelectLayer 0-7 dimensions 0-7 dimensions No Yes NA
IShapeLayer 1 or more dimensions 1 dimension No No NA
IShuffleLayer 0-7 dimensions 0-7 dimensions No No No
ISliceLayer 1-7 dimensions 1-7 dimensions No No Yes
ISoftMaxLayer 1-7 dimensions 1-7 dimensions No No Yes
ITopKLayer 1-7 dimensions
  • Output1: 1-7 dimensions
  • Output2: 1-7 dimensions
Yes No Yes
ITripLimitLayer 0 dimensions has no outputs No No NA
IUnaryLayer 1-7 dimensions 1-7 dimensions No No No

For more information about each of the TensorRT layers, see TensorRT Layers.

3. Layers And Precision

The section lists the TensorRT layers and the precision modes that each layer supports. It also lists the ability of the layer to run on Deep Learning Accelerator (DLA).
For more information about additional constraints, see DLA Supported Layers.

For more information about each of the TensorRT layers, see TensorRT Layers. To view a list of the specific attributes that are supported by each layer, refer to the TensorRT API documentation.

Table 3. List of supported precision modes per TensorRT layer.
Layer FP32 FP16 INT8 INT32 DLA FP16 DLA INT8
IActivationLayer Yes Yes Yes No Yes3 Yes4
IConcatenationLayer Yes Yes Yes Yes Yes5 Yes5
IConstantLayer Yes Yes Yes Yes No No
IConvolutionLayer > 2D Convolution Yes Yes Yes No Yes Yes
IConvolutionLayer > 3D Convolution Yes Yes Yes No No No
IDeconvolutionLayer > 2D Deconvolution Yes Yes Yes No Yes Yes6
IDeconvolutionLayer > 3D Deconvolution Yes Yes No No No No
IDequantizeLayer No No Yes No No No
IElementWiseLayer Yes Yes No Yes Yes7 Yes8
IFillLayer Yes No No Yes No No
IFullyConnectedLayer Yes Yes Yes No Yes Yes
IGatherLayer Yes Yes No Yes No No
IIdentityLayer Yes Yes Yes Yes No No
IIteratorLayer Yes Yes No Yes No No
ILoopOutputLayer Yes Yes No Yes No No
IPluginV2Layer Yes Yes Yes No No No
ILRNLayer Yes Yes Yes No Yes No
IMatrixMultiplyLayer Yes Yes No No No No
IPaddingLayer Yes Yes Yes No No No
IParametricReluLayer Yes Yes Yes No No No
IPoolingLayer > 2D Pooling Yes Yes Yes No Yes9 Yes9
IPoolingLayer > 3D Pooling Yes Yes No No No No
IQuantizeLayer Yes No No No No No
IRaggedSoftMaxLayer Yes No No No No No
IRecurrenceLayer Yes Yes No Yes No No
IReduceLayer Yes Yes No No No No
IResizeLayer Yes Yes No No No No
IRNNv2Layer Yes Yes No No No No
IScaleLayer Yes Yes Yes No Yes10 Yes10
ISelectLayer Yes Yes No Yes No No
IShapeLayer11 Yes Yes Yes Yes No No
IShuffleLayer Yes Yes Yes Yes No No
ISliceLayer Yes Yes No12 Yes No No
ISoftMaxLayer Yes Yes No No No No
ITopKLayer Yes Yes No No No No
ITripLimitLayer Yes Yes No Yes No No
IUnaryLayer Yes Yes No No No No
Note: DLA with FP16/INT8 precision with some restrictions on layer parameters.

4. Hardware And Precision

The following table lists NVIDIA hardware and which precision modes each hardware supports. TensorRT supports all NVIDIA hardware with capability SM 5.0 or higher. It also lists the availability of Deep Learning Accelerator (DLA) on this hardware. Refer to the following tables for the specifics.
Note: Support for CUDA Compute Capability version 3.0 has been removed. Support for CUDA Compute Capability versions below 5.0 may be removed in a future release and is now deprecated.
Table 4. Supported hardware
CUDA Compute Capability Example Device TF32 FP32 FP16 INT8 FP16 Tensor Cores INT8 Tensor Cores DLA
8.6 NVIDIA A10 Yes Yes Yes Yes Yes Yes No
8.0 NVIDIA A100/GA100 GPU Yes Yes Yes Yes Yes Yes No
7.5 Tesla T4 No Yes Yes Yes Yes Yes No
7.2 Jetson AGX Xavier No Yes Yes Yes Yes Yes Yes
7.0 Tesla V100 No Yes Yes Yes Yes No No
6.2 Jetson TX2 No Yes Yes No No No No
6.1 Tesla P4 No Yes No Yes No No No
6.0 Tesla P100 No Yes Yes No No No No
5.3 Jetson TX1 No Yes Yes No No No No
5.2 Tesla M4 No Yes No No No No No
5.0 Quadro K2200 No Yes No No No No No

Deprecated hardware

Table 5. List of supported precision mode per hardware.
CUDA Compute Capability Example Device FP32 FP16 INT8 FP16 Tensor Cores INT8 Tensor Cores DLA
3.7 Tesla K80 Yes No No No No No
3.5 Tesla K40 Yes No No No No No

Removed hardware

Table 6. List of supported precision mode per hardware.
CUDA Compute Capability Example Device FP32 FP16 INT8 FP16 Tensor Cores INT8 Tensor Cores DLA
3.0 Tesla K10 Yes No No No No No

5. Software Versions Per Platform

The section lists the supported software versions based on platform.
Table 7. List of supported platforms per software version.
  Compiler version Python versions
Ubuntu 16.04 x86-64 gcc 8.3.1 3.5
Ubuntu 18.04 x86-64 gcc 8.3.1 3.6
Ubuntu 20.04 x86-64 gcc 8.3.1 3.8
CentOS 7.9 x86-64 gcc 8.3.1 3.6
CentOS 8.3 x86-64 gcc 8.3.1 3.8
SLES 15 x86-64 gcc 8.3.1 N/A
Windows 10 x64 MSVC 2017u5 N/A
CentOS 8.3 ppc64le Clang 10.0.1 3.8
Ubuntu 20.04 SBSA gcc 8.4.0 3.8
JetPack AArch64 gcc 7.5.0 3.6
Note: Python versions supported when using Debian or RPM packages. When using Python wheel files, versions 3.5, 3.6, 3.7, 3.8, and 3.9 are supported.

6. Supported Ops

The section lists the operations that are supported in a Caffe or TensorFlow framework and in the ONNX TensorRT parser.

ONNX

Since the ONNX parser is an open source project, the most up-to-date information regarding the supported operations can be found here.

These are the ONNX operators that are supported by TensorRT:
  • Abs
  • Acos
  • Acosh
  • And
  • Asin
  • Asinh
  • Atan
  • Atanh
  • Add
  • ArgMax
  • ArgMin
  • AveragePool
  • BatchNormalization
  • Cast
  • Ceil
  • Celu
  • Clip
  • Concat
  • Constant
  • ConstantOfShape
  • Conv
  • ConvTranspose
  • Cos
  • Cosh
  • CumSum
  • DepthToSpace
  • DequantizeLinear
  • Div
  • Dropout
  • Elu
  • Equal
  • Erf
  • Exp
  • Expand
  • EyeLike
  • Flatten
  • Floor
  • Gather
  • GatherElements
  • Gemm
  • GlobalAveragePool
  • GlobalLpPool
  • GlobalMaxPool
  • Greater
  • GreaterOrEqual
  • GRU
  • HardSigmoid
  • Identity
  • ImageScaler
  • InstanceNormalization
  • LeakyRelU
  • Less
  • LessOrEqual
  • Log
  • LogSoftmax
  • Loop
  • LpNormalization
  • LpPool
  • LRN
  • LSTM
  • MatMul
  • Max
  • MaxPool
  • Mean
  • Min
  • Mul
  • Neg
  • Not
  • Or
  • Pad
  • ParametricSoftplus
  • Pow
  • PRelu
  • QuantizeLinear
  • RandomUniform
  • RandomUniformLike
  • Range
  • Reciprocal
  • ReduceL1
  • ReduceL2
  • ReduceLogSum
  • ReduceLogSumExp
  • ReduceMax
  • ReduceMean
  • ReduceMin
  • ReduceProd
  • ReduceSum
  • ReduceSumSquare
  • Relu
  • Reshape
  • Resize
  • ReverseSequence
  • RNN
  • ScaledTanh
  • Scan
  • Selu
  • Shape
  • Sigmoid
  • Sin
  • Sinh
  • Size
  • Slice
  • Softmax
  • SoftmaxCrossEntropyLoss
  • Softplus
  • Softsign
  • SpaceToDepth
  • Split
  • Sqrt
  • Squeeze
  • Sub
  • Sum
  • Tan
  • Tanh
  • ThresholdedRelu
  • Tile
  • TopK
  • Transpose
  • Unsqueeze
  • Upsample
  • Where

Caffe

These are the operations that are supported in a Caffe framework:
  • BatchNormalization
  • BNLL
  • Clip13
  • Concatenation
  • Convolution
  • Crop
  • Deconvolution
  • Dropout
  • ElementWise
  • ELU
  • InnerProduct
  • Input
  • LeakyReLU
  • LRN
  • Permute
  • Pooling
  • Power
  • Reduction
  • ReLU, TanH, and Sigmoid
  • Reshape
  • SoftMax
  • Scale

TensorFlow

These are the operations that are supported in a TensorFlow framework:
  • Add, AddV2, AddN, Sub, Mul, Div, FloorDiv, RealDiv, Minimum, Maximum
  • AvgPool, AvgPool3D14
  • ArgMin
  • AvgPool
  • BiasAdd
  • Cast14
  • Clip
  • CombinedNonMaxSuppression
  • ConcatV2
  • Const
  • Conv2D, Conv3D
  • Conv2DBackpropInput, Conv3DBackpropInputV2
  • ConvTranspose2D
  • DepthToSpace
  • DepthwiseConv2dNative
  • Elu
  • ExpandDims
  • FusedBatchNorm, FusedBatchNormV2, FusedBatchNormV3
  • FusedConv2DBiasActivation
  • GatherV2
  • Identity
  • LeakyReLU
  • MatMul, BatchMatMul, BatchMatMulV2
  • MaxPool, MaxPool3D14
  • Mean
  • Negative, Abs, Sqrt, Recip, Rsqrt, Pow, Exp, Log, Square
  • Pad is supported if followed by one of these TensorFlow layers: Conv2D, DepthwiseConv2dNative, MaxPool, and AvgPool.
  • Pack, Unpack
  • ReLU, TanH, Sigmoid
  • Relu6
  • Reshape
  • ResizeBilinear, ResizeNearestNeighbor
  • Sin, Cos, Tan, Asin, Acos, Atan, Sinh, Cosh, Asinh, Acosh, Atanh, Ceil, Floor
  • Selu
  • Shape14
  • Slice, StridedSlice
  • SoftMax
    Note: If the input to a TensorFlow SoftMax op is not NHWC, TensorFlow will automatically insert a transpose layer with a non-constant permutation, causing the UFF converter to fail. It is therefore advisable to manually transpose SoftMax inputs to NHWC using a constant permutation.
  • Softplus
  • Softsign
  • SpaceToDepth
  • Split
  • SquaredDifference
  • Squeeze
  • TopKV2
  • Transpose

For the list of ops supported in UFF, see UFF Operators.

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.

ARM

ARM, AMBA and ARM Powered are registered trademarks of ARM Limited. Cortex, MPCore and Mali are trademarks of ARM Limited. All other brands or product names are the property of their respective holders. "ARM" is used to represent ARM Holdings plc; its operating company ARM Limited; and the regional subsidiaries ARM Inc.; ARM KK; ARM Korea Limited.; ARM Taiwan Limited; ARM France SAS; ARM Consulting (Shanghai) Co. Ltd.; ARM Germany GmbH; ARM Embedded Technologies Pvt. Ltd.; ARM Norway, AS and ARM Sweden AB.

OpenCL

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

Trademarks

NVIDIA, the NVIDIA logo, and cuBLAS, CUDA, CUDA Toolkit, cuDNN, DALI, DIGITS, DGX, DGX-1, DGX-2, DGX Station, DLProf, GPU, JetPack, Jetson, Kepler, Maxwell, NCCL, Nsight Compute, Nsight Systems, NVCaffe, NVIDIA Ampere GPU architecture, NVIDIA Deep Learning SDK, NVIDIA Developer Program, NVIDIA GPU Cloud, NVLink, NVSHMEM, PerfWorks, Pascal, SDK Manager, T4, Tegra, TensorRT, TensorRT Inference Server, Tesla, TF-TRT, Triton Inference Server, Turing, and Volta are trademarks and/or registered trademarks of NVIDIA Corporation in the United States and other countries. Other company and product names may be trademarks of the respective companies with which they are associated.

1 This build supports CUDA compute capability 8.6. It is compatible with CUDA 11.1 and CUDA 11.2. User-mode driver compatible with the runtime CUDA version is required and >= 465 is suggested for best performance.
2 These CUDA versions are supported using a single build, built with CUDA Toolkit 11.3 update 1. It is compatible with all CUDA 11.x (x <= 3) versions and only requires driver 450.x. For future CUDA 11.x (x > 3) versions, the corresponding 11.x driver is required which matches the CUDA Toolkit.
3 Partial support. Yes for ReLU, Clipped ReLU, Leaky ReLU, Sigmoid and TanH activation types only.
4 Partial support. Yes for ReLU, Clipped ReLU, Leaky ReLU, Sigmoid and TanH activation type only.
5 Partial support. Yes for concatenation across C dimension only.
6 Partial support. Yes for ungrouped deconvolutions and No for grouped.
7 Partial support. Yes for sum, sub, prod, min and max elementwise operations only.
8 Partial support. Yes for sum elementwise operation only.
9 Partial support. Yes for max and average padding inclusive pooling type only.
10 Partial support. DLA does not support power on scale layer.
11 Output is always INT32.
12 Partial support. Yes for unstrided Slice and No for strided.
13 When using the Clip operation, Caffe users must serialize their layers using ditcaffe.pb.h instead of caffe.pb.h in order to import the layer into TensorRT.
14 Supported only in TensorFlow 2.0.