Tesla Compute Cluster (TCC)

NVIDIA® Nsight™ Application Development Environment for Heterogeneous Platforms, Visual Studio Edition 5.5 User Guide
Send Feedback


For most GPUs, you do not have to do anything specific in NVIDIA Nsight to enable debugging on a Tesla Compute Cluster (TCC) device. You don't have to modify your Visual Studio project or enable any specific setting. The TCC device simply shows up as a standard CUDA device. For some GPUs, the default mode is not TCC. See below for more information.

Do not kill a process that is executing code on a TCC device and paused on a breakpoint, except through the normal Stop Debugging command (SHIFT+F5) in Visual Studio. Abnormal termination of a target application that is paused during a debugging session on a TCC device results in unpredictable behavior. It causes future calls to cuCtxInit() to hang indefinitely, even though the killed process seems to terminate normally. The only way to recover is to reboot the target machine. See the latest release notes to learn about any changes in this behavior.

Limitations

There are 2 main limitations you will encounter when debugging code running on a TCC device:

Setting TCC Mode for Tesla Products

NVIDIA GPUs can exist in one of two modes: TCC or WDDM. TCC mode disables Windows graphics and is used in headless configurations, whereas WDDM mode is required for Windows graphics. NVIDIA GPUs also come in three classes:

NVIDIA Nsight Graphics debugging requires the target GPU be in WDDM mode.

NVIDIA Nsight Compute debugging does not require a specific driver mode.

The NVIDIA Control Panel will show you what mode your GPUs are in; alternately, you can use the nvidia-smi command to generate a table that will display your GPUs and what mode they are using.

To change the TCC mode, use the NVIDIA SMI utility. This is located by default at C:\Program Files\NVIDIA Corporation\NVSMI. Use the following syntax to change the TCC mode:

nvidia-smi -g {GPU_ID} -dm {0|1}

0 = WDDM

1 = TCC

NOTE: NVIDIA® Nsight™ Visual Studio Edition does not support CUDA debugging in TCC mode on Pascal Family GPUs. Support will be added in an upcoming release.

About TCC

The TCC (Tesla Compute Cluster) driver is a Windows driver that supports CUDA C/C++ applications. The driver enables remote desktop services, and reduces the CUDA kernel launch overhead on Windows. Note that the TCC driver disables graphics on the Tesla products.

The main purpose of TCC and the Tesla products is to aid applications that use CUDA to perform simulations, and large scale calculations (especially floating-point calculations), such as image generation for professional use and scientific fields of study.

The benefits of using the Tesla Compute Cluster driver package:

The TCC driver was specifically designed to be used with Microsoft's Windows HPC Server 2008. However, NVIDIA's TCC driver can be used with operating systems other than Windows HPC Server 2008. The NVIDIA TCC driver does not have the same pinned allocation limits or memory fragmentation behavior as WDDM. You can mix TCC drivers with XP-style display drivers.

For more information about supported operating systems, and compatibility with other NVIDIA drivers, refer to the documentation on NVIDIA Tesla:

http://www.nvidia.com/object/tesla_computing_solutions.html

For more information about NVIDIA hardware compatibility on Windows HPC Server 2008, see:

http://technet.microsoft.com/en-us/library/ff793340(WS.10).aspx

To search the NVIDIA web site for Tesla drivers, see:

http://www.nvidia.com/Download/index.aspx


NVIDIA® GameWorks™ Documentation Rev. 1.0.180319 ©2014-2018. NVIDIA Corporation. All Rights Reserved.