Installing a second K20 card.

Accelerated Computing CUDA CUDA Setup and Installation
1

Hi there,

I look after a server that is used for GPU computing by the Department of Statistics at the University of Warwick.

The machine in question currently has SuSE Enterprise linux 11 X64 and has a single Tesla K20c card it it.

The department has just purchased a second card to be fitted to this machine, an Identical Tesla K20.

Whilst I am confident about being able to install the hardware will there be any software configuration that I will be required to do once the new card (and upgraded PSUs) is fitted?

Also is there a way that I can test the functionality / presence of the two cards once the physical upgrade is complete.

Cheers.

Phill.

2

Assuming the system (hardware/BIOS) can support 2 cards, you shouldn’t need to do anything software wise.

After installing the 2nd card and rebooting, some simple tests to verify the functionality of the second card might be to run:

nvidia-smi

and verify that both cards are reported.

then run the deviceQuery cuda sample code, and it should report both cards.