NVLink SLI to use two RTX2080 Ti's as one GPU to get a total memory of 22GB for CUDA 9.X

Dear Nvidia support

I’d like to use NVLink SLI for cuda, while I don’t use tensor.

Do NVLink-two-RTX2080’s with a new driver, e.g. 430.50 work as one GPU to get a total memory of 22GB by for Cuda 9.1 or 9.1?

OS is CentOS 7.2 and 7.5

Best wishes

YuK

P.S. This was first questioned to Mansoor at nvhelp@nvidia.com, who replied

"Please contact our Linux support team about your questions or query from the below mentioned link and they will be able to assist you further.

https://devtalk.nvidia.com/default/board/98/"