Language Selection

English French German Italian Portuguese Spanish

Vulkan: SIGGRAPH 2019 News and NVIDIA Focus

Filed under
Graphics/Benchmarks
  • NVIDIA Continues To Be Involved With Making Vulkan More Appropriate For Machine Learning

    NVIDIA engineers continue to be among those in the Vulkan technical sub-group working to advance machine learning for this API.

    Vulkan machine learning is being worked on for functionality like NVIDIA's DLSS, bots, character animations, and other functionality that can be tailored to machine learning in high frame-rate applications. There's also the benefit of Vulkan being an industry standard unlike CUDA and friends.

  • NVIDIA 435.17 Linux Beta Driver Adds Vulkan + OpenGL PRIME Render Offload

    NVIDIA this morning introduced their 435 Linux driver series currently in beta form with the release of the 435.17 Linux build. With this new driver comes finally the best PRIME/multi-GPU support they have presented to date.

    The NVIDIA 435.17 driver has a new PRIME render offload implementation supported for Vulkan and OpenGL (with GLX). This PRIME offloading is about using one GPU for display but having the actual rendering be done on a secondary GPU, as is common with many of today's high-end notebooks that have Intel integrated graphics paired with a discrete NVIDIA GPU.

  • Vulkan Video Decoding Coming In H1'2020, Ray-Tracing Progressing

    The Khronos Group has posted their material from the SIGGRAPH 2019 graphics conference and includes some interesting updates on Vulkan and their ongoing efforts.

    In addition to making Vulkan better for machine learning, ray-tracing and video decode are two other topics of interest to us.

Liam Dawe's coverage

  • NVIDIA have released the 435.17 beta driver with Vulkan and OpenGL support for PRIME render offload

    NVIDIA have a little present available for Linux fans today, with the release of the 435.17 beta driver now being available.

    This is a beta driver and it includes quite the highlight with the addition of PRIME render offload support for Vulkan and OpenGL. This is where you might have your Intel GPU running most normal applications, with an NVIDIA chip then powering your games. It's usually found in Notebooks and it's been a source of annoyance for NVIDIA Notebook owners for a long time, so it's really pleasing to see proper progress like this.

NVIDIA 435.17 Linux Beta Driver Adds PRIME Offloading For Vulkan

  • NVIDIA 435.17 Linux Beta Driver Adds PRIME Offloading For Vulkan and OpenGL+GLX

    The latest NVIDIA 435.17 Linux beta driver has added Vulkan and OpenGL+GLX support for PRIME render offload.

    PRIME is a collection of features in the Linux kernel, display server, and various drivers to enable GPU offloading with multi-GPU configurations under Linux, like laptops using NVIDIA Optimus (which use an integrated Intel GPU and a discrete NVIDIA GPU).

    Thanks to the new on-demand PRIME render offload, you'll be able to run specific application on the discrete NVIDIA GPU, while using the integrated GPU for everything else, saving battery power.

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.

More in Tux Machines

How to compile a Linux kernel in the 21st century

In computing, a kernel is the low-level software that handles communication with hardware and general system coordination. Aside from some initial firmware built into your computer's motherboard, when you start your computer, the kernel is what provides awareness that it has a hard drive and a screen and a keyboard and a network card. It's also the kernel's job to ensure equal time (more or less) is given to each component so that your graphics and audio and filesystem and network all run smoothly, even though they're running concurrently. The quest for hardware support, however, is ongoing, because the more hardware that gets released, the more stuff a kernel must adopt into its code to make the hardware work as expected. It's difficult to get accurate numbers, but the Linux kernel is certainly among the top kernels for hardware compatibility. Linux operates innumerable computers and mobile phones, embedded system on a chip (SoC) boards for hobbyist and industrial uses, RAID cards, sewing machines, and much more. Read more

Life with an offline laptop

When I think about an offline laptop, I immediately think I will miss IRC, mails, file synchronization, Mastodon and remote ssh to my servers. But do I really need it _all the time_?

As I started thinking about preparing an old laptop for the experiment, differents ideas with theirs pros and cons came to my mind.

Read more

today's leftovers

  • Huawei mulls open-source chip design if US ban continues

    Huawei Technologies Co said Friday that it would consider using RISC-V, an open-source chip architecture, if the US government continues restricting its access to the latest technologies from the UK firm ARM Holdings for a long time. Xu Zhijun, rotating chairman of Huawei, said in an interview in Shenzhen, Guangdong province, that the company has already obtained the perpetual license to ARM's V8 architecture technology, so the US government ban does not affect its current launch schedule of chips. "If ARM's new technologies are not available in the future, we can also use RISC-V, an architecture which is open to all companies. The challenge is not insurmountable," Xu said.

  • From Spark To Airflow And Presto: Demystifying The Fast-Moving Cloud Data Stack

    Putting data to work starts with exploration, or inspecting data so that you know what you have to work with and its characteristics. Presto is excellent for exploring large, unstructured data sets because it uses storage efficiently, which keeps costs down, and it’s compatible with SQL, a language data analysts are familiar with.  Spark, on the other hand, is great for exploring data sets when programming is required, such as being able to manipulate data for use in data science or machine learning. It has good support for non-SQL interfaces.

  • Databricks launches AutoML Toolkit for model building and deployment

    Databricks today introduced its AutoML Toolkit, an automated end-to-end machine learning service made to accommodate developers with a range of experience.

  • DigitalOcean Managed Databases add MySQL, Redis support

    DigitalOcean Managed Databases introduced support for open source relational database MySQL and in-memory database Redis to eliminate the complexity involved in managing, scaling and securing database infrastructure. DigitalOcean, a cloud computing vendor offering infrastructure-as-a-service platforms for software developers, intends its new managed database offerings to enable developers to focus more exclusively on building apps and boosting productivity.

  • How Storj Is Building a Storage Cloud Without Owning a Single Disk

    Led by Docker's former CEO, the startup is crowdsourcing empty disk space from desktops and data centers around the world.

  • HPC Computing Is Replacing Supercomputers In Enterprise: Jeff Reser

    Jeff Reser – Global Product and Solutions Marketing Manager of SUSE talks about High-Performance Computing.

  • Mable & The Wood is a fairly unusual Metroidvania out now with Linux support

    Featuring a sweet fairy-powered protagonist wielding a sword so big they can hardly move, Mable & The Wood certainly presents a healthy amount of charm. Developed by Triplevision Games, a solo outfit from the UK, with publishing from Graffiti Games. "Mable is a passion project for me and for so long I worked on it by myself," said Andrew Stewart, Founder of Triplevision Games. “Thanks to Graffiti, I was able to have additional support to release the game sooner and on multiple platforms. Players on Steam can finally get their hands on the brilliant title today, and fear not Switch and Xbox One players, that version will be releasing very soon."

Programming: Python Shows, Golang and GNOME/GLib Work

  • Python Bytes: #144 Are you mocking me? It won't work!
  • Talk Python to Me: #226 Building Flask APIs for data scientists

    If you're a data scientist, how do you deliver your analysis and your models to the people who need them? A really good option is to serve them over Flask as an API. But there are some special considerations you might keep in mind. How should you structure this API? What type of project structures work best for data science and Flask web apps? That and much more on this episode of Talk Python To Me with guest AJ Pryor.

  • Golang or go home: how Curve is taking Golang to new heights

    Emerging only in 2009, Golang is still relatively new and not as widely used as other mainstream coding languages. This young language was incubated inside Google, and has already been proven to perform well on a massive scale. We wanted to share with you a few reasons why we love Golang (Go) and how Curve is using it. Go has excellent characteristics for scalability and services written using it typically have very small memory footprints. Because code is compiled into a single static binary, services can also be containerised with ease, making it much simpler to build and deploy. These attributes make Go an ideal choice for companies building microservices, as you can easily deploy into a highly available and scalable environment such as Kubernetes. Go has everything you need to build APIs as part of its standard library.

  • GTimeVal deprecation in GLib 2.61.2

    One of the latest changes in GLib (released in 2.61.2) is the deprecation of GTimeVal, g_get_current_time(), and a number of other time functions. This is because we can’t guarantee they’re wide enough on all platforms to be year-2038-safe. Instead, you should use GDateTime or, if you just need to store epoch time, guint64. They are year-2038-safe — and with that, GLib should be entirely year-2038-safe. GTimeVal is used in a number of places, and widespread (but simple) changes will need to be made to stop using it. You will likely have already seen some deprecation warnings popping up to inform you of this, if you use any C-based and GLib-based libraries. If you can’t allocate time to fixing these deprecation warnings yet, you can silence them by explicitly stating your minimum and maximum supported versions of GLib. If your minimum supported version of GLib is older than 2.62, you won’t see deprecation warnings for GTimeVal (since it was deprecated in 2.62, and your code is claiming to need to support older GLib versions than that).

  • Mayank Sharma: GSoC’19 - GVfs and the Google Backend demystified

    Note: Due to time limitations, I haven’t been able to devote much time to writing a blog post. Each time I started, some or the other thing bothered me and I ended up having a draft. My humble apologies to my readers. So, over the past 3 months or so, I’ve been working on the Google Backend for GVfs (GNOME Virtual File System), and as of today, the backend is in a state where it’s completely useable. Earlier, a large number of operations were disabled. So, if you tried to copy a file from one folder to the other, you’d be given an error “Operation not supported”. Now, you may be wondering what’s there in a simple copy operation that the developers/maintainers can’t fix, or shouldn’t something like Google Drive backend for GVfs receive better attention since a great deal of peope keep their important data on their G-Drive? The answer isn’t a yes or no, and it’s much more subjective since it pertains to the state of current open-source software. One of the big reasons has been that OSS always lacks man-power, and that the problem at hand wasn’t trivial in any sense. My mentor (Ondrej Holy), is the sole maintainer of a project as big as GVfs, and he certainly doesn’t have the time of look at each backend’s issues.