we got a some notice using Nsight Debbugger with Tesla P100

Hi everyone.
we’re developing finance programs in CUDA with Tesla P100.
but, we have some problem about nsight debbug.
we got a some message while debbug in VS2015 with Nsight.

message:

" A CUDA context was created on a GPU that is not currently debuggable.
Breakpoints will be disabled.
"

what can we do something wrong?

To enable others to speculate more effectively, you might want to mention

(1) What CUDA version you are running
(2) The version of the installed driver package
(3) The Nsight version being used

Are there other GPUs installed in the machine besides the P100?

  1. We are running cuda8
  2. Driver version is 23.21.13.8508
  3. The nsight version is 5.2.0.16321

Our develop server spec is

  1. Intel XEON CPU E5-2667 v4@3.20GHZ
  2. 512 GB physical memory
  3. Window server 2016 standard
  4. Tesla P100-PCIE-16GB X 6EA

I do not recognize this as a valid Windows driver version number. If this is a fairly recent driver it should be something like 385.xx. Maybe for Windows Server a different driver versioning scheme is used? I don’t know.

There is probably enough information now that someone who has deeper insights into Nsight (pun intended) can make a reasonable guess as to what might be going on. I don’t have any idea why you would be getting this error message, as the relevant hardware and software seems to be in place, and I assume you can run CUDA applications in general (just not debug them with Nsight).

windows driver versions are reported somewhat differently depending on where you look for the driver version (e.g. device manager vs. display driver control panel). This looks to me like it is probably a 385.08 driver.

when you run

nvidia-smi -a

from a command line, what is the driver model reported? Is it WDDM or TCC ? (need to know for all 6 GPUs)

are you connecting to this machine using RDP or some kind of remoting technology?

It’s all TCC.
and, it’s exactly right your guess.
We’re developing on RDP.