Jump to content
LaptopVideo2Go Forums

GeForce 452.39 Quadro ODE / Data center


Recommended Posts

Release 450 is an 'Optimal Drivers for Enterprise' (ODE) branch release. ODE branches are designed and tested to provide long-term stability and availability, making drivers from these branches ideal for enterprise customers and other users requiring application and hardware certification.

Quadro ODE drivers are a superset of the NVIDIA Studio Drivers and provide all the benefits of the Studio Driver, in addition to Quadro-specific enhancements and testing.

What’s New in Version 450 U4

Fixed Issues in Version 450 U4

  • [PTC Creo]: Models may get corrupted due to wrong VBO index buffer used.
  • [OptiX][Quadro RTX]: Performance degradation may occur with apps when using Optix.

Open Issues in Version 450 U4

  • [Alchemist]: The application crashes when either changing the display resolution or switching to another display with a different resolution.
  • To workaround, avoid changing the resolution, or disconnecting from or connecting to an external display with the application opened.


Link to comment
Share on other sites

FYI: the Tesla GPGPUs which are no longer called Tesla and never should have been named Tesla in first place, will ne named "Data Center" in future 🤪


Version 450.80.02(Linux)/452.39(Windows)

Release notes for the Release 450 family of NVIDIA® Data Center GPU Drivers for Linux and Windows.

This edition of Release Notes describes the Release 450 family of NVIDIA® Data Center GPU Drivers for Linux and Windows. NVIDIA provides these notes to describe performance improvements, bug fixes and limitations in each documented version of the driver.

1. Version Highlights

This section provides highlights of the NVIDIA Data Center GPU R450 Driver (version 450.80.02 Linux and 452.39 Windows).

For changes related to the 450 release of the NVIDIA display driver, review the file "NVIDIA_Changelog" available in the .run installer packages.

Driver release date: 09/30/2020

1.1. Software Versions

07/28/2020: For this release, the software versions are listed below.

  • CUDA Toolkit 11: 11.03

    Note that starting with CUDA 11, individual components of the toolkit are versioned independently. For a full list of the individual versioned components (e.g. nvcc, CUDA libraries etc.), see the CUDA Toolkit Release Notes

  • NVIDIA Data Center GPU Driver: 450.80.02 (Linux) / 452.39 (Windows)

  • Fabric Manager: 450.80.02 (Use nv-fabricmanager -v)


    • (NVIDIA A100 SKU200 with heatsink for HGX A100 8-way and 4-way)

    • (NVIDIA A100 SKU202 w/o heatsink for HGX A100 4-way)

  • NVSwitch VBIOS:

  • NVFlash: 5.641

Due to a revision lock between the VBIOS and driver, VBIOS versions >= must use corresponding drivers >= 450.36.01. Older VBIOS versions will work with newer drivers.

For more information on getting started with the NVIDIA Fabric Manager on NVSwitch-based systems (for example, HGX A100), refer to the Fabric Manager User Guide.

1.2. Improvements

  • Added a new system-level interface using nvidia-capabilities via /dev for Multi-Instance GPU (MIG) on NVIDIA A100. For more information on these new device node capabilities, refer to the MIG user guide.

  • Added support for CUDA compatibility for CUDA 11.1, where CUDA 11.1 applications will work on supported GPUs when used with drivers >= 450.80.00 on Linux. Refer to CUDA Compatibility for more information.

1.3. Fixed Issues

  • Various security issues were addressed. For additional details on the med-high severity issues, review the NVIDIA Security Bulletin 5075 .

  • Fixed an issue where, when NVLink training encounters a fatal error, link training is still considered successful resulting in broken nodes assigned to clients. An escape case in the NVLink link state reporting caused the driver to report the links as "Active" and functional even though the link had received a fatal error. This fatal error does not allow traffic to be sent over the link, and thus would cause P2P apps to fail. This is a reporting issue, but not a functionality issue. The lacking reporting case was addressed, so the failing link state is properly reported.

  • Fixed a software race condition that caused link training failure (CUDA_ERROR_LAUNCH_FAILED). A race condition within the driver polling code caused a false failure in some of the link training steps. This causes the driver to not continue link training even though training actually was successful up to that point. The fix corrects these sequences so that these false failures no longer happen.

Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

  • Create New...