Dynamic Parallelism with a GeForce GTX 750 Ti

Accelerated Computing CUDA CUDA Programming and Performance
1

Hello,

I have a GeForce GTX 750 Ti and want to do Dynamic Parallelism with it ; however, when my code runs, NSight desactivates breakpoints due to preemption mode.
I desactivated preemption mode, it still didn’t work, and I figured out I should make my GPU headless.

The problem is, I don’t have the needed second GPU, and my motherboard doesn’t support two GPUs at the same time anyway.

My questions are :

  • What kind of GPU can I use to display, alongside the headless 750 Ti ?
  • What kind of motherboard would support it ?

Please help, thanks in advance !

2

Wait, was your conclusion that the only way to get dynamic parallelism is to stop your GPU from being your display GPU as well?

3

TomTomCanNot,

If you cannot install two GPUs in the system and you want to debug a CUDA application using CUDA Dynamic Parallelism you will likely have to run Nsight VSE on a second computer and remote debug the system with the 750Ti.

Greg

4

Thank you, I will try this !

5

Greg

do we have a pointer to howto 2xGPU usage of Nsight with CUDA Dynamic Parallelism on a single box?

GTX 750Ti would be good, do we have a reference or example motherboard or box?

presumably GTX 980 also need 2x as getting pricey…

Jonathan