【Linux】CUDA Toolkit和cuDNN版本对应关系（更新至2022年6月，附官网永久更新链接）

目录

一、前言

NVIDIA® CUDA® Toolkit 为创建高性能 GPU 加速应用程序提供了一个开发环境。借助 CUDA 工具包，您可以在 GPU 加速的嵌入式系统、桌面工作站、企业数据中心、基于云的平台和 HPC 超级计算机上开发、优化和部署您的应用程序。该工具包包括 GPU 加速库、调试和优化工具、C/C++ 编译器以及用于部署应用程序的运行时库。
全球的深度学习研究人员和框架开发人员都依赖cuDNN来实现高性能GPU加速。它使他们可以专注于训练神经网络和开发软件应用程序，而不必花时间在底层GPU性能调整上。
本篇博客记录了了CUDA Toolkit 和cuDNN的版本对应关系，并给出了下载官网的地址，一些初学者可能需要这份版本对应关系表，下载之前可能需要注册一个账号。

二、对应关系

• cuDNN官网下载地址：https://developer.nvidia.com/rdp/cudnn-archive
• CUDA 工具包下载地址：https://developer.nvidia.com/cuda-toolkit-archive
• 官网可查对应关系：
  cuDNNCUDAcuDNN v8.4.0 (April 1st, 2022)CUDA 11.xcuDNN v8.4.0 (April 1st, 2022)CUDA 10.2cuDNN v8.3.3 (March 18th, 2022)CUDA 11.5cuDNN v8.3.3 (March 18th, 2022)CUDA 10.2cuDNN v8.3.2 (January 10th, 2022)CUDA 11.5cuDNN v8.3.2 (January 10th, 2022)CUDA 10.2cuDNN v8.3.1 (November 22nd, 2021)CUDA 11.5cuDNN v8.3.1 (November 22nd, 2021)CUDA 10.2cuDNN v8.3.0 (November 3rd, 2021)CUDA 11.5cuDNN v8.3.0 (November 3rd, 2021)CUDA 10.2cuDNN v8.2.4 (September 2nd, 2021)CUDA 11.4cuDNN v8.2.4 (September 2nd, 2021)CUDA 10.2cuDNN v8.2.2 (July 6th, 2021)CUDA 11.4cuDNN v8.2.2 (July 6th, 2021)CUDA 10.2cuDNN v8.2.1 (June 7th, 2021)CUDA 11.xcuDNN v8.2.1 (June 7th, 2021)CUDA 10.2cuDNN v8.2.0 (April 23rd, 2021)CUDA 11.xcuDNN v8.2.0 (April 23rd, 2021)CUDA 10.2cuDNN v8.1.1 (Feburary 26th, 2021)CUDA 11.0,11.1 and 11.2cuDNN v8.1.1 (Feburary 26th, 2021)CUDA 10.2cuDNN v8.1.0 (January 26th, 2021)CUDA 11.0,11.1 and 11.2cuDNN v8.1.0 (January 26th, 2021)CUDA 10.2cuDNN v8.0.5 (November 9th, 2020)CUDA 11.1cuDNN v8.0.5 (November 9th, 2020)CUDA 11.0cuDNN v8.0.5 (November 9th, 2020)CUDA 10.2cuDNN v8.0.5 (November 9th, 2020)CUDA 10.1cuDNN v8.0.4 (September 28th, 2020)CUDA 11.1cuDNN v8.0.4 (September 28th, 2020)CUDA 11.0cuDNN v8.0.4 (September 28th, 2020)CUDA 10.2cuDNN v8.0.4 (September 28th, 2020)CUDA 10.1cuDNN v8.0.3 (August 26th, 2020)CUDA 11.0cuDNN v8.0.3 (August 26th, 2020)CUDA 10.2cuDNN v8.0.3 (August 26th, 2020)CUDA 10.1cuDNN v8.0.2 (July 24th, 2020)CUDA 11.0cuDNN v8.0.2 (July 24th, 2020)CUDA 10.2cuDNN v8.0.2 (July 24th, 2020)CUDA 10.1cuDNN v8.0.1 RC2 (June 26th, 2020)CUDA 11.0cuDNN v8.0.1 RC2 (June 26th, 2020)CUDA 10.2cuDNN v7.6.5 (November 18th, 2019)CUDA 10.2cuDNN v7.6.5 (November 5th, 2019)CUDA 10.1cuDNN v7.6.5 (November 5th, 2019)CUDA 10.0cuDNN v7.6.5 (November 5th, 2019)CUDA 9.2cuDNN v7.6.5 (November 5th, 2019)CUDA 9.0cuDNN v7.6.4 (September 27, 2019)CUDA 10.1cuDNN v7.6.4 (September 27, 2019)CUDA 10.0cuDNN v7.6.4 (September 27, 2019)CUDA 9.2cuDNN v7.6.4 (September 27, 2019)CUDA 9.0cuDNN v7.6.3 (August 23, 2019)CUDA 10.1cuDNN v7.6.3 (August 23, 2019)CUDA 10.0cuDNN v7.6.3 (August 23, 2019)CUDA 9.2cuDNN v7.6.3 (August 23, 2019)CUDA 9.0cuDNN v7.6.2 (July 22, 2019)CUDA 10.1cuDNN v7.6.2 (July 22, 2019)CUDA 10.0cuDNN v7.6.2 (July 22, 2019)CUDA 9.2cuDNN v7.6.2 (July 22, 2019)CUDA 9.0cuDNN v7.6.1 (June 24, 2019)CUDA 10.1cuDNN v7.6.1 (June 24, 2019)CUDA 10.0cuDNN v7.6.1 (June 24, 2019)CUDA 9.2cuDNN v7.6.1 (June 24, 2019)CUDA 9.0cuDNN v7.6.0 (May 20, 2019)CUDA 10.1cuDNN v7.6.0 (May 20, 2019)CUDA 10.0cuDNN v7.6.0 (May 20, 2019)CUDA 9.2cuDNN v7.6.0 (May 20, 2019)CUDA 9.0cuDNN v7.5.1 (April 22, 2019)CUDA 10.1cuDNN v7.5.1 (April 22, 2019)CUDA 10.0cuDNN v7.5.1 (April 22, 2019)CUDA 9.2cuDNN v7.5.1 (April 22, 2019)CUDA 9.0cuDNN v7.5.0 (Feb 25, 2019)CUDA 10.1cuDNN v7.5.0 (Feb 21, 2019)CUDA 10.0cuDNN v7.5.0 (Feb 21, 2019)CUDA 9.2cuDNN v7.5.0 (Feb 21, 2019)CUDA 9.0cuDNN v7.4.2 (Dec 14, 2018)CUDA 10.0cuDNN v7.4.2 (Dec 14, 2018)CUDA 9.2cuDNN v7.4.2 (Dec 14, 2018)CUDA 9.0cuDNN v7.4.1 (Nov 8, 2018)CUDA 10.0cuDNN v7.4.1 (Nov 8, 2018)CUDA 9.2cuDNN v7.4.1 (Nov 8, 2018)CUDA 9.0cuDNN v7.3.1 (Sept 28, 2018)CUDA 10.0cuDNN v7.3.1 (Sept 28, 2018)CUDA 9.2cuDNN v7.3.1 (Sept 28, 2018)CUDA 9.0cuDNN v7.3.0 (Sept 19, 2018)CUDA 10.0cuDNN v7.3.0 (Sept 19, 2018)CUDA 9.0cuDNN v7.2.1 (August 7, 2018)CUDA 9.2cuDNN v7.1.4 (May 16, 2018)CUDA 9.2cuDNN v7.1.4 (May 16, 2018)CUDA 9.0cuDNN v7.1.4 (May 16, 2018)CUDA 8.0cuDNN v7.1.3 (April 17, 2018)CUDA 9.1cuDNN v7.1.3 (April 17, 2018)CUDA 9.0cuDNN v7.1.3 (April 17, 2018)CUDA 8.0cuDNN v7.1.2 (Mar 21, 2018)CUDA 9.1 & 9.2cuDNN v7.1.2 (Mar 21, 2018)CUDA 9.0cuDNN v7.0.5 (Dec 11, 2017)CUDA 9.1cuDNN v7.0.5 (Dec 5, 2017)CUDA 9.0cuDNN v7.0.5 (Dec 5, 2017)CUDA 8.0cuDNN v7.0.4 (Nov 13, 2017)CUDA 9.0cuDNN v6.0 (April 27, 2017)CUDA 8.0cuDNN v6.0 (April 27, 2017)CUDA 7.5cuDNN v5.1 (Jan 20, 2017)CUDA 8.0cuDNN v5.1 (Jan 20, 2017)CUDA 7.5cuDNN v5 (May 27, 2016)CUDA 8.0cuDNN v5 (May 12, 2016)CUDA 7.5cuDNN v4 (Feb 10, 2016)CUDA 7.0 and later.cuDNN v3 (September 8, 2015)CUDA 7.0 and later.cuDNN v2 (March 17,2015)CUDA 6.5 and later.cuDNN v1 (cuDNN 6.5 R1)

  三、如何使用

安装流程参考：

• 【Linux】安装CUDA 11.2 和 cuDNN 8.4.0并检查是否安装成功

四、总结

• CUDA Toolkit 和cuDNN的版本对应关系很重要，是正确利用GPU计算资源的关键。如果本文能给你带来帮助的话，点个赞鼓励一下作者吧！

五、参考

[1] cuDNN 官网下载地址：https://developer.nvidia.com/rdp/cudnn-archive
[2] CUDA 工具包下载地址：https://developer.nvidia.com/cuda-toolkit-archive