Hello everyone,
I’m working on my first CUDA project however I’ve hit a small roadbump.
I’m running the latest CUDA toolkit SDK and drivers on an nForce 780a SLI Motherboard which I am currently using the 8300mGPU to render my desktop.
However, when I run a cuda application it only lists the mGPU and not the 8800GT that I have.
I’ve read that applications can only run for a short time if they are run on the GPU that runs your desktop however, I need this application to run for about a 5 minutes to generate some graph results, so what do I do?
Applications can run as long as you want (even hours or days). A single kernel execution can’t take more than a couple of seconds on a desktop rendering GPU. Most often you will launch fast kernels in a loop.
To select a device different than your primary display adapter, use cudaSetDevice(int deviceNumber), where deviceNumber 0 = primary card, 1 = second…
To all: I forgot to mention my specs and dev env
Its a Asus Crosshair 2 Formula
with 8GB of ram running vista 64bit
a few 500GB hds and a 8800GT that I’m using specificly for development. My company uses VS2008 but I was able to obtain a copy of VS2005.
However, I’m building a server to run code on a Tesla 1070 when funds allow.
Back to the problem,
The problem is that my 8800GT isn’t showing up as a CUDA device. The getting started manual referred to an NVIDIA desktop, does that mean that I need a display attached to the 8800 to get everything going?
Are you running on windows Vista? If so, read the release notes, it mentions the limitations of running multiple CUDA cards in Vista: each one needs a monitor and a desktop extended to it if I recall correctly.