Nexus FAQ!

From Nvidia Knowledge base:
What features are included in the Nexus Beta 1 build?

Nexus Beta 1 includes the following components:

Debugger
Support for debugging CUDA C and HLSL 10 source directly on the GPU hardware.
Analyzer
Trace now supports CUDA C, DX, and OpenGL API calls in addition to CUDA C kernel executions and memory transfers.
Graphics Inspector
Supports the debugging and profiling of Direct3D 10 frames, including frame breakdown using draw call scrubbing, texture and render target inspection.

What features can only used remotely over TCP/IP in Nexus Beta 1?

Some features of Nexus (like the source Debugger) can only be used remotely over TCP/IP because they will full halt the GPU on which your code is being debugged. Use the following table to see which features should only be used remotely:


Local Single machine with a single GPU Remote Two machines, connected by TCP/IP
Debugger
For CUDA C and HLSL No* Yes
Analyzer

For CUDA C, DirectX, and OpenGL Yes Yes
Graphics Inspector
D3D Frame Debugger and Profiler Yes Yes


* Nexus Beta 1 contains experimental local CUDA C Debugger support on a machine with multiple GPUs.

How do I setup and use Nexus Beta 1's experimental support for debugging CUDA C programs locally on a single machine with multiple GPUs?

Nexus Beta 1 has experimental support for debugging CUDA C programs locally on a single machine. Please remember that under certain circumstances, debugging locally can cause stability issues on your machine, and that remote debugging over TCP/IP is the only officially supported way to debug your CUDA C code.
For best results, please note the following requirements and setup. There is additional information on these topics contained in the Nexus User Guide, which is installed with the Nexus build.
Hardware Requirements:

Your computer must have multiple GPUs. At least one GPU must be G92 or GT200 based. This GPU is what you must run your application's kernels on. Nexus Preview 2 contains no special support for ensuring that your kernels are running on the correct GPU - it is your application's responsibility.
System Requirements:

1. For best results, set-up your computer to run "Headless" on the Debug GPU. Running "headless" means that your Windows desktop will not be extended to one of your GPUs (and that GPU will be used to run the code you wish to debug).Instructions for doing this are provided in the Nexus User Guide.
2. Disable WPF hardware acceleration on the your computer. Instructions are provided in the Nexus User Guide.
Application Requirements:
1. Your application must not run kernels on the same GPU as the one running your computer's display. If your application runs kernels on the display GPU and stops at a breakpoint, your entire system will likely experience stability issues.
2. Your application must not use Direct3D or OpenGL in addition to CUDA C.
Nexus Setup:

1. Once your hardware and application setup is complete, you can use the Nexus > Start CUDA Debugging... menu item to start debugging CUDA C programs normally.

2. Make sure that you set the debug target (Nexus User Properties when you right-click on the project) to localhost, and that you have the Nexus Monitor installed and running.

What version of Visual Studio do I need in order to run Nexus?

You will need: Visual Studio 2008 SP1 Standard, Professional or Team System
Visual Studio Express is not supported because Express does not support VS extensions and add-ins.