Language Selection

English French German Italian Portuguese Spanish

Intel and AMD in Linux: Intel Uncore Frequency, Per-Client Engine Busyness, Navi GPUs

Filed under
  • Intel Uncore Frequency Driver On Linux Is Closer To Mainline With Latest Patches

    The Intel Uncore Frequency driver for Linux allows reading and setting the uncore frequency, which controls the RING / Last Level Cache (LLC) clocks. Increasing the uncore frequency can help with improving memory latency or at least making the latency consistent by avoiding the dynamic uncore frequency selection. Increasing the clocks obviously come with increased power/heat. By default, the Intel uncore frequency is dynamic based upon the CPU performance state and power constraints.

    Intel has long published as part of their data sheets that the uncore frequency is controlled via a specific MSR (register 0x620) while now the intel_uncore frequency driver makes it easy to read and manipulate it from user-space. Interested server administrators could already manipulate the MSR as desired, but the Intel Uncore Frequency Linux driver makes it very easy now to read and set via new sysfs interfaces.

  • Intel Revs Linux Patches For Per-Client Engine Busyness - Allowing For Great GPU Insight

    One of the set of patches for Intel's Linux kernel graphics driver that have been floating around for more than one year is about exposing per-client (process) statistics in how each application is making use of the GPU's render/blitter/video hardware and various insightful statistics related to that. The patches aren't queued for mainline yet but at least a new revision of the work was published.

  • RADV's Next-Gen Geometry Code Continues To Be Revised For Navi GPUs

    The NGG (Next-Gen Geometry) support with Navi continues to be refined by the open-source AMD Linux graphics drivers with the RADV Vulkan driver seeing a fresh batch of fixes/clean-ups, inspired in part by the NGG code from the RadeonSI and AMDVLK drivers.

    The latest batch of RADV NGG work comes via Valve open-source driver developer Samuel Pitoiset. Landing in Mesa 20.0-devel today is a performance optimization, support for the NGG passthrough mode, disabling of vertex grouping, and other tweaks.

    Nothing too major itself but the latest in a long series of NGG activity for the open-source AMD Linux graphics drivers.

More in Tux Machines

RHEL no-cost* vs openSUSE Leap

Ever since Red Hat announced that they are changing the development model of CentOS and making it an upstream project rather than downstream, it left many CentOS users frowning. No matter what argument brought forward, CentOS users, especially running production machines, relied on the stability of an enterprise-grade Linux distribution. Compiled from RHEL sources, CentOS offered such stability that it powered many web servers and enjoyed a massive 20% share of the top 500 supercomputers of the world. Read more

10 of the Best Linux Debuggers for Software Engineers

Debuggers are essential for locating bugs in programs. There is a plethora of robust Linux debuggers that make it easy to find weak points in your applications. We will outline some of these applications in this guide. Try some of these tools to get a feel of how debugging works in Linux. Read more

Snapcraft GNOME Extension Update

Snaps are confined software packages for Linux. They were originally designed / intended for IoT use cases so are optimised for size, bundling dependencies, are compressed on disk and auto update. They can also be used to package server software, like NextCloud, and desktop software like Signal Desktop. There’s millions of desktops, routers, servers and other interesting devices with snaps installed. There’s a bunch of common components that snap publishers started bundling in their snaps which bloated them out a bit. Snaps have had (for some time) a concept of “shared content” such that one snap may consume assets from another snap. The reason we use the hand-wavy term “assets” and “content” is because while it could be binary programs and libraries which are shared between snaps, it’s not just limited to that. A theme or bundle of themes can be shared too. Read more

Konsole Re-Flow Lines

One day I was looking at the MR (Merge Request) and saw Tomaz Canabrava’s sketch showing the terminal re-flow lines while it shrinks, and I just thought it would be great to have it fully working. The first thing to have a line re-flow is define how to mark a line with “continues in the next line”. This is the most important thing, otherwise you can’t go back to original state. My first thought was to set a next line char with something not printable, and then the first screen re-flow on both ways prototype was done. To improve speed, and hold lines before send to memory, the next thing I did was change the _screenLines holder from an array to a QVector type. It was an improvement in speed, specially to re-flow. No need to a new memory allocation, no copy and no delete, it was just update the QVector content and send from the QVector to history, if needed, and resize it. Read more