NVIDIA News Archives

Last week NVIDIA open-sourced 12k lines of C header files for Blackwell GPUs to help in the open-source driver efforts, namely for Nouveau / NVK and the in-development NOVA Rust driver. On Friday they made public some additional header files for helping in the Blackwell and Hopper open-source driver enablement.

NVIDIA announced this week that they are bringing their CUDA software to RISC-V processors.

Similar to the 3D class header files previously open-sourced by NVIDIA for prior generation GPUs, yesterday NVIDIA carried out a similar open-source move to publish all the 3D class header files for their newest Blackwell graphics processors.

NVIDIA software engineers ended out the week by releasing a new beta of their RTX Neural Texture Compression (NTC) SDK. The RTXNTC software is NVIDIA's interesting solution for compressing material texture sets with very promising results for helping to reduce game data sizes moving forward.

Merged today for the LLVM/Clang compiler is -mcpu=gb10 support for catering to NVIDIA's forthcoming Grace Blackwell GB10 Superchip.

NVIDIA previously warned CUDA users that CUDA 12.x is the last for Maxwell, Pascal, and Volta GPUs. NVIDIA overnight now officially confirmed that the Maxwell / Pascal / Volta GPU support is going to end in their Linux driver with the upcoming NVIDIA R580 Linux driver series.

NVIDIA today released a new version of their R575 series stable Linux driver to provide a few additional fixes for customers.

The open-source, unofficial VA-API driver implementation for NVIDIA GPUs is out with a new release. The NVIDIA-VAAPI-Driver project continues to build VA-API support atop NVIDIA's NVDEC interface to allow the NVIDIA proprietary driver stack to work with VA-API-only applications like the Mozilla Firefox web browser on Linux.

NVIDIA just released their v575.57.08 driver as the first stable Linux driver in their R575 release branch.

Last year NVIDIA began publicly outlining known Wayland limitations for their driver and future support plans. This week NVIDIA put out updated information concerning known gaps in Wayland support as well as their ongoing future driver plans for enhancing the feature compatibility under this X11 successor.

NVIDIA CUDA 12.9 is now available as the newest minor feature update to NVIDIA's GPU compute stack. CUDA 12.9 adds compiler targetr support for SM 10.3 and 12.1, compiler support for "family-specific architectures", new NVML counters being exposed, and other minor feature improvements. The NVIDIA CUDA 12.9 documentation is also now more verbose in encouraging anyone still relying on Maxwell, Pascal, and Volta hardware to upgrade.

Merged for the Linux 6.15 kernel were the very early boilerplate code around the NOVA driver as a new, open-source and Rust-written NVIDIA Linux kernel graphics/display driver. This successor to the Nouveau kernel driver is going to leverage the NVIDIA GPU System Processor (GSP) to make it easier to develop and maintain this open-source driver. But depending upon the GSP also means the NOVA driver will only work with RTX 20 class GPUs and newer. This driver is going to be built up gradually within the mainline Linux kernel and coming out this Easter were a new set of 16 patches for further laying the NOVA groundwork.

With today's launch of the GeForce RTX 5060 Ti graphics card, NVIDIA has also introduced their first Linux driver beta in the R575 driver series.

One of many Linux kernel patch series that NVIDIA has been working on recently to upstream to the mainline Linux kernel are the adjustments needed for Linux running on their newest Smart Switch networking products for the data center.

As a win for the open-source community from NVIDIA, the company recently announced they are making their PhysX and Flow GPU-accelerated source code open-source.

The GCC and LLVM Clang open-source compilers have landed support for the NVIDIA Olympus cores for NVIDIA's Vera CPU that is part of their next-gen Rubin microarchitecture succeeding Blackwell.

Haiku OS developer X512 has managed a rather impressive feat: porting NVIDIA's open-source kernel modules to Haiku. Not only did he get NVIDIA's official Linux kernel modules running on Haiku but he also ported the Mesa NVK Vulkan driver to be able to run atop the NVIDIA kernel driver interface.

In addition to NVIDIA's GTC conference going on right now, they also released today an updated R570 Linux driver build.

NVIDIA's RTX Remix software for remastering classic games with a variety of visual enhancements reached version 1.0 today with some additional big improvements.

For quite a while Red Hat engineers have been developing the open-source, Rust-written NOVA driver to in effect serve as the successor to the reverse-engineered Nouveau driver that isn't too actively developed in more recent times. But unlike Nouveau's extensive range of NVIDIA GPU support, the NOVA driver is intentionally limited to the RTX 20 "Turing" GPUs and newer where there is the NVIDIA GPU System Processor (GSP) with the firmware support to leverage for an easier driver-writing experience. The very initial NOVA driver code was sent out on Sunday for DRM-Next ahead of the Linux 6.15 merge window.

It's not only AMD that is working on Vulkan/SPIR-V support for machine learning / AI software but NVIDIA has been working on improvements too for enhancing Vulkan-powered machine learning software. The outlook for using Vulkan within machine learning software is quite positive and even able to offer similar performance to NVIDIA's prized CUDA.

NVIDIA engineers closed out February by releasing the NVIDIA 570.123.01 Vulkan beta driver for Linux and on the Windows side was the NVIDIA 572.63 driver release.

Going into beta just under one month ago was the NVIDIA 570.86.16 Linux driver that brought initial support for GeForce RTX 50 "Blackwell" graphics cards as well as Vulkan driver improvements, Variable Rate Refresh (VRR) support with multiple displays, GPU overclocking exposed by default, and various other refinements to this official NVIDIA Linux driver. Out today is the NVIDIA 570.124.04 stable Linux driver release.

NVIDIA on Thursday published their first public beta of their RTX Neural Texture Compression "RTX NTC" software development kit.

Merged last year for the Linux 6.12 kernel was sched_ext for allowing extensible scheduler possibilities by allowing schedulers to be implemented as eBPF code and dynamically loaded into the kernel. This allows for rapidly developing new schedulers as well as exploring other new possibilities around more intelligent kernel scheduling decisions. Meta, Google, Canonical (Ubuntu), and others have been big proponents of sched_ext and NVIDIA is also increasingly vocalizing their support for these extensible scheduler opportunities.

The open-source Rust CUDA project has been "rebooted" to get back onto the effort of allowing NVIDIA CUDA compute kernels to be coded within the Rust programming language.

The NVIDIA 570.86.16 beta Linux driver was just published in time for the GeForce RTX 5080 and RTX 5090 graphics cards hitting store shelves this morning.

With now having a Linux driver for running GPU compute workloads on the GeForce RTX 5090 (as mentioned, Linux gaming benchmarks will come following the formal R570 Linux driver release in the coming days that is better optimized for gaming), I ran some additional GPU compute benchmarks on the GeForce RTX 5090 "Blackwell" graphics card over the weekend.

Now that NVIDIA is rolling out the "Blackwell" GPU driver support, it looks like the NVIDIA Maxwell, Pascal, and Volta generations will soon be moving to a legacy driver branch.

Following the unboxing embargo earlier this week for the NVIDIA GeForce RTX 5090, today the review embargo lifts for the GeForce RTX 5090 Founders Edition.

The latest from a rather active CES 2025 is NVIDIA announcing the GeForce RTX 50 "Blackwell" series line-up.

NVIDIA compiler engineers have spent the past several months working on a proposed GCC option -flto-partition=locality for having the compiler optimize the code layout for locality between callees and callers as part of the link-time optimization (LTO) process. For some workloads NVIDIA is finding this -flto-partition=locality compiler option being of significant help for bettering the CPU performance.

This year NVIDIA's official Linux graphics driver enjoyed much more robust Wayland support, their open-source kernel modules have matured greatly and are now being used by default, and their proprietary Vulkan and OpenGL drivers remain in good standing for performant Linux gaming and workstation graphics. NVIDIA's Linux driver stack had a rather great year.

NVIDIA today announced the Jetson Orin Nano Super Developer Kit as their "most affordable generative AI Supercomputer" with this upgraded Jetson Nano offering 1.7x better GenAI performance while also costing less than its predecessor. This new product looks like an exciting addition to the NVIDIA Jetson line-up and will have performance benchmarks soon on Phoronix.

NVIDIA published EGL-Wayland 1.1.17 on Monday as the newest update to this Wayland EGL external platform library to provide client-side Wayland support to EGL atop the EGLDevice and EGLStream extensions.

For the past month and a half the NVIDIA R565 Linux driver series has been in public beta with a number of (X)Wayland improvements, DMA-BUF enhancements, VKD3D fixes, and a variety of other enhancements. Today the NVIDIA 565.77 Linux driver was released as the first stable build in the series.

NVIDIA's RTX Remix software for remastering DirectX 8 and DirectX 9 era games is out with the newest version of the RTX-Remix runtime that is powered in part by DXVK for Direct3D to Vulkan mapping.

The NVIDIA MLX5 driver for NVIDIA Mellanox ConnectX-5 network adapters is preparing to introduce a new Data Direct Placement "DDP" feature with the upcoming Linux 6.13 kernel merge window.

Longtime NVIDIA Linux engineer Aaron Plattner shared a status update on Friday around the current feature parity difference between the NVIDIA driver stack on X11 and under (X)Wayland.

Going back to 2016 we've known of NVIDIA beginning to use RISC-V to replace their Falcon micro-controller and other micro-controllers within their graphics processors to using this common open-source ISA. That use has continued to grow and an unofficial estimate now puts it at around one billion RISC-V cores shipping in 2024 NVIDIA chips.

The first NVIDIA R565 series Linux driver beta was released this morning in the form of the NVIDIA 565.57.01 driver release.

Building off the existing Linux support for GPU Direct RDMA / Peer-To-Peer DMA functionality, a set of patches were posted by NVIDIA today enabling this P2P DMA support to also work for device-private pages.

In addition to NVIDIA engineers being at XDC 2024 in Montreal last week for talking about their Wayland driver plans, there was also a presentation by NVIDIA's Daniel Dadap around current Linux challenges in supporting dynamic display mux hardware on modern laptops with iGPU/dGPU combinations and their hopes for improving the support.

At the X.Org Developer's Conference (XDC 2024) happening this week in Montreal, NVIDIA shared a road-map around their Wayland plans as well as encouraging Wayland compositors to target the Vulkan API.

The NVGRACE-GPU VFIO driver was introduced for handling Virtual Function I/O support with the NVIDIA Grace Hopper Superchip so that the GPU device could be assigned to guests using KVM/QEMU and similar for virtualization. The NVGRACE-GPU driver is now being extended for supporting the forthcoming NVIDIA Grace Blackwell "GB" designs.

While open-source enthusiasts like to criticize NVIDIA for not maintaining upstream, in-tree kernel graphics driver support (though things have been changing there), for other areas of their vast hardware portfolio they are much better upstream Linux kernel citizens and often at the forefront of new driver innovations. One of the leading examples of that is around the NVIDIA Mellanox networking driver support. With Linux 6.12 they've landed a new feature that has been described as "a sign of things to come, I think we will see more of this in the next 10 years."

NVIDIA engineers have sent out an exciting set of Linux kernel patches for enabling NVIDIA vGPU software support for virtual GPU support among multiple virtual machines (VMs). In aiming for upstream-focused Linux support, this NVIDIA vGPU support is built around the adapted Nouveau driver with the code previously posted for splitting up the Nouveau/NVKM driver components.

OpenBMC as the Linux Foundation project backed by vendors like Intel / Microsoft / Google / Meta for an open-source BMC firmware stack continues to be a growing success. This alternative to long-used proprietary BMC software stacks continues to grow in popularity with AMD now using it on their reference motherboards and Supermicro being another notable user with some of their server platforms. Not entirely new but been meaning to write about it and NVIDIA talked more openly about it this week: NVIDIA is also a big supporter and user of OpenBMC for their high-end AI/HPC servers and BlueField DPU hardware.

Building off the prior NVIDIA 560 beta driver releases, the NVIDIA 560.35.03 stable Linux driver was released today for providing the latest official NVIDIA graphics/compute support for Linux systems.

There's a new release of NVIDIA's EGL-Wayland project for an EGL External Library Platform library implementing EGL on top of EGLDevice and EGLStream extensions. This week's NVIDIA EGL-Wayland 1.15 release is primarily centered on delivering Wayland explicit sync fixes.