On Thu, Jan 5, 2017 at 2:15 PM, Susan Chacko <susanc.helix.nih.gov> wrote:
> Ah, someone suggested backchannel that I should try setting the
> CUDA_VISIBLE_DEVICES to ensure that 2 GPUs were being used.
>
> deviceQuery says there are 2 devices called 0 and 1
> % ./deviceQuery -noprompt | egrep "^Device"
> Device 0: "Tesla K80"
> Device 1: "Tesla K80"
>
> but nvidia-smi says there are 4.
>
​deviceQuery is implemented using the CUDA API. nvidia-smi is not. So if
you want to know what pmemd.cuda (and any other CUDA application) will
find, use deviceQuery.
In fact, device numbers in nvidia-smi will not even necessarily match the
device numbers reported by deviceQuery. This is actually true on my
machine which has 2 different CUDA-capable cards and deviceQuery and
nvidia-smi have opposite assignments for devices 0 and 1:
swails.Batman ~ $ deviceQuery | grep "^Device"
Device 0: "GeForce GTX 680"
Device 1: "GeForce GT 740"
swails.Batman ~ $ nvidia-smi | grep GeForce
| 0 GeForce GT 740 Off | 0000:01:00.0 N/A |
N/A |
| 1 GeForce GTX 680 Off | 0000:07:00.0 N/A |
N/A |
HTH,
Jason
--
Jason M. Swails
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Thu Jan 05 2017 - 13:00:02 PST
