Multiple NVIDIA drivers

Accelerated Computing CUDA CUDA Setup and Installation
1

Is it possible to have multiple NVIDIA drivers existing simultaneously ?

Currently, I have two C2070, C2075s :

$ cat /proc/driver/nvidia/version
NVRM version: NVIDIA UNIX x86_64 Kernel Module 295.59 Wed Jun 6 21:19:40 PDT 2012
GCC version: gcc version 4.4.6 20110731 (Red Hat 4.4.6-3) (GCC)

I would like to also install this driver ? Is it possible ?
NVIDIA-Linux-x86_64-340.32

This is my config
$ nvidia-smi
Thu Sep 25 11:56:16 2014
±-----------------------------------------------------+
| NVIDIA-SMI 3.295.59 Driver Version: 295.59 |
|-------------------------------±---------------------±---------------------+
| Nb. Name | Bus Id Disp. | Volatile ECC SB / DB |
| Fan Temp Power Usage /Cap | Memory Usage | GPU Util. Compute M. |
|===============================+======================+======================|
| 0. Tesla C2070 | 0000:02:00.0 On | 0 0 |
| 36% 85 C P8 N/A / N/A | 3% 177MB / 5375MB | 0% Default |
|-------------------------------±---------------------±---------------------|
| 1. Tesla C2070 | 0000:03:00.0 Off | 0 0 |
| 32% 83 C P8 N/A / N/A | 0% 9MB / 5375MB | 0% Default |
|-------------------------------±---------------------±---------------------|
| 2. Tesla C2075 | 0000:83:00.0 Off | 0 0 |
| 30% 60 C P12 36W / 225W | 0% 9MB / 5375MB | 0% Default |
|-------------------------------±---------------------±---------------------|
| 3. Tesla C2075 | 0000:84:00.0 Off | 0 0 |
| 30% 61 C P12 29W / 225W | 0% 9MB / 5375MB | 0% Default |
|-------------------------------±---------------------±---------------------|
| Compute processes: GPU Memory |
| GPU PID Process name Usage |
|=============================================================================|
| No running compute processes found |
±----------------------------------------------------------------------------+

2

No it’s not possible. If you attempt to install the newer driver, it will remove and replace the older one.

The newer driver should be compatible with any older CUDA toolkits that you want to use, if that is your reason for wanting to retain the older driver.