Logo Questions Linux Laravel Mysql Ubuntu Git Menu
 

Does cuDNN library works with All nvidia graphic cards?

I study the use of cuDNN library in my project. But my nvidia graphic card is little bit old.

I search on the net if cuDNN works with all graphic cards. I didn,t find even in their principal page.

Which nvidia graphic cards are compatible with cuDNN?

like image 780
ProEns08 Avatar asked May 26 '16 13:05

ProEns08


2 Answers

Kepler and newer, ie compute capability 3.0 and above.

https://developer.nvidia.com/cudnn

Supported on Windows, Linux and MacOS systems with Pascal, Kepler, Maxwell, Tegra K1 or Tegra X1 GPUs.

like image 146
Robert Crovella Avatar answered Nov 06 '22 06:11

Robert Crovella


The deep learning SDK documentation also(1) contains relevant information on the requirements to run cuDNN.

There you will find a link to the list of current GPUs and their compute capabilities, from which you can select the graphic card having the required compute capability >=3.0 (2):

https://developer.nvidia.com/cuda-gpus

Here is an excerpt for GeForce cards: enter image description here

Wikipedia has this even more convenient CUDA supported GPU table listing most (if not all) NVDIA graphic cards with their compute capabilities (the most up-to-date information is however likely to be found from the NVDIA link on cuda-gpus above).

(1) in reference to Robert Crovella's answer

(2) as of May 2018

like image 32
calocedrus Avatar answered Nov 06 '22 05:11

calocedrus