Language Selection

English French German Italian Portuguese Spanish

Graphics/Benchmarks

GCC 10 Link-Time Optimization Benchmarks On AMD Threadripper

Filed under
Graphics/Benchmarks

Stemming from the recent news in Fedora 32 potentially LTO'ing packages by default for better performance and not yet having checked on the Link-Time Optimization performance of the in-development GCC 10, here is a fresh look at the possible performance gains from making use of link-time optimizations for generating faster binaries. This round of testing was done on the AMD Ryzen Threadripper 3960X and is complementary to the recent Profile Guided Optimization benchmarks.

Read more

Kernel: Linux 5.6 and Latest of Nouveau

Filed under
Graphics/Benchmarks
Linux
  • Linux 5.6 Is Going To Be A Prominent Kernel With Features From USB4 To WireGuard

    While there still is several weeks to go until the Linux 5.5 kernel reaches stable and that marking the start of the Linux 5.6 merge window, already from the work we've been tracking in the various "next" branches, this first full kernel cycle of 2020 is going to be a big one. 

  • NVIDIA Sent Out Some Fresh Nouveau Patches Just Before Christmas

    Shortly before Christmas were a couple open-source Nouveau driver patches volleyed by NVIDIA. Some of that work is now queuing in the Nouveau DRM tree ahead of the Linux 5.6 merge window.

    Some of the recent Nouveau work courtesy of NVIDIA engineers include generalizing the NV Block Linear DRM format and supporting NVIDIA format modifiers in atomic mode-setting blobs. Those go along with work by NVIDIA's James Jones with Mesa patches in then exposing EGL's EXT_transition_format_modifier support. As Jones explained, "This allows differentiating between surfaces compatible with Tegra GPUs and desktop GPUs, improved performance when using the compressed layouts, and differentiating between formats supported by nvc0-class hardware and other GPU families which use slightly different variations of the block linear buffer layout."

AMD Ryzen 5 1600 AF CPU Review & Benchmarks: Best CPU Under $100 in 2020

Filed under
Graphics/Benchmarks

This isn’t a revisit of the old AMD Ryzen 5 1600 – it’s a review of the new variant, named the AMD Ryzen 5 1600 “AF” by the community, dubbed as such for its SKU change from AE to AF. The AMD R5 1600 AF is a brand new CPU with an old, old name from 2017. It’s mostly an R5 2600, in that it’s a slower variant of the Zen+ CPU from the 2000-series, but with a 1000-series name. AMD silently released the 1600 AF as an $85 option, but it’s on 12nm instead of 14nm and carries other 2nd-Gen Ryzen features. In today’s review of the new $85 processor, we’ll look at performance versus the original R5 1600, the R5 2600, and overclocking performance, since a 12nm 1600 AF should do about the same OC as a 12nm Ryzen 2000 part, which were typically 100-200MHz higher than the 1000-series.

The R5 1600 AF is a weird, weird refresh. It’s mostly odd that AMD didn’t just name it Ryzen 3 3300X or Ryzen 5 3550. They already have the 3000 family with Zen+ architecture and the 3000G with Zen1 architecture, so it wouldn’t dilute the naming and it’d be a much more successful, higher selling product with a lot of media fanfare. Instead, it just sounds like a two-year-old part, but it’s really not. We can’t fault AMD for its naming and it doesn’t particularly bother us, it’s just a bit odd from a marketing standpoint. Maybe AMD doesn’t want to sell a lot of these.

Read more

Kernel and Graphics: The Linux Kernel Code of Conduct Committee, Zink and X11/X.Org Server

Filed under
Graphics/Benchmarks
Linux
  • The Linux Kernel Code of Conduct Committee Begins Offering Regular Reports

    Greg Kroah-Hartman on Friday released a December 2019 report on the behalf of the Code of Conduct committee. This marks of what they hope will be regular (monthly?) reports from the committee on their work.

    For last month they received just one report simply described as "insulting behavior in email." Stemming from an investigation, that lone report resulted in "education and coaching" but without shining any further light onto the incident.

    At least a lone report for the month on the kernel mailing list that sees thousands of posts per month (from a quick count on their archive, around 9,760 mailing list posts for December).

    Previously from the start of the CoC in September 2018 through July of 2019 they indicated three instances of insulting behavior in emails and one case of inappropriate language in the kernel source tree. From August through November there were no reports. In the four earlier reports, they all resulted in the "education and coaching."

  • Zink Is Moving Closer To OpenGL 3.0 Support Over Vulkan

    Zink was one of the Mesa/Gallium3D innovations that saw mainline status in 2019 for offering OpenGL support atop Vulkan hardware drivers. While an interesting approach, so far only the dated OpenGL 2.1 support has been exposed but the Collabora-led effort is closing in on OpenGL 3.0 capabilities.

    Zink could play a vital role in the future when GPU hardware vendors begin focusing on Vulkan support exclusively or at least otherwise dropping OpenGL support for future generations of hardware. Though the primary limitations for this generic OpenGL-over-Vulkan layer is that Zink is quite slow compared to dedicated hardware drivers and that only OpenGL 2.1 has been exposed to date.

  • X.Org Server Development Hits A Nearly Two Decade Low

    With Red Hat shifting their support to Wayland and expecting the X.Org Server to go into a hard maintenance mode quickly, in 2019 indeed it did.

    Without any other companies investing significantly into the X.Org Server itself with engineering resources, the X.Org Server is seeing little work these days beyond work to XWayland for running X11 applications on top of Wayland and also work to GLAMOR as the 2D acceleration code over OpenGL. It's rare seeing activity elsewhere sans the occasional commits to the xf86-video-modesetting DDX driver.

Linux 5.4 vs. Liquorix Kernel Benchmarks For AMD Ryzen + Radeon Gaming On Ubuntu

Filed under
Graphics/Benchmarks

The Liquorix kernel is the long-standing effort for providing a "better distro kernel" optimized for desktop/multimedia/gaming workloads. As it's been a while since last testing the Liquorix kernel spin of Linux, I recently carried out some tests of its Linux 5.4 based kernel compared to Ubuntu's generic mainline PPA images of Linux 5.4 as well as the low-latency kernel variety.

The Liquorix kernel continues to tweak its Linux kernel configuration and add in extra patches for aiming to optimize the kernel for gaming/desktop-type workloads. Liquorix employs the MuQSS scheduler, Zen interactive tuning, a 1000Hz tick rate, hard kernel preemption, BFQ for the I/O scheduler, and other tweaks.

On an AMD Ryzen 9 3900X + Radeon RX Vega 64 system I ran benchmarks with its 5.4.0-6.3-liquorix official kernel build for Ubuntu compared to the Linux 5.4.6 mainline kernel builds both of the generic and lowlatency binaries from the Ubuntu Mainline Kernel PPA.

Read more

Graphics: Radeon Gallium3D, Libinput 1.15 and Mesa

Filed under
Graphics/Benchmarks
  • Radeon Gallium3D Fixes Up 10-Bit HEVC Video Decode Support

    Hitting Mesa 20.0-devel a short time ago were a set of patches to the Radeon Gallium3D video code for fixing 10-bit HEVC video decode support.

    The AMD Radeon open-source Linux graphics driver has supported 10-bit HEVC video decode acceleration using the P016 format, but that has caused problems with GStreamer-based software while working fine for the likes of MPV and FFmpeg.

  • libinput 1.15.0
    libinput 1.15 is now available.
    
    No significant changes over the RC, merely quirks for the Logitech MX Master
    2S and the Madcatz R.A.T.3 mouse. Thus, to fill the void with useful
    stuff, let's copy-paste the RC1 announcement text here so it looks like
    we've been busy:
    
    A few new features, a lot of maintenance work. On the feature front we have
    scroll button locking and tablet pad keys support.
    
    Scroll button locking holds a scroll button logically down even when the
    button is physicall released. This applies to button scrolling (hold the
    button while moving a mouse up/down generates scroll events). Previously the
    button had to be physicall held down during the scroll motion which is
    difficult for a number of users. When the scroll button lock is enabled, the
    first click of the button holds it logically down, the second click releases
    it. This feature comes with the usual set of configuration hooks:
           libinput_device_config_scroll_set_button_lock()
           libinput_device_config_scroll_get_button_lock()
           libinput_device_config_scroll_get_default_button_lock()
    
    Some tablet devices have hardware buttons that are designed to map to a
    specific functionality (e.g. to open the OSD). These buttons are now
    supported through the new LIBINPUT_EVENT_TABLET_PAD_KEY event. Noteworthy
    here is that for those events the key code matters (unlike the tablet pad
    button code which is just a sequential number). See the documentation for
    more details. In addition to the event we have a new API to query whether a
    given tablet has a specific hw key:
        libinput_device_tablet_pad_has_key()
    
    
    Other than those, a whole bunch of device specific-fixes and general
    cleanups. Touchpad scrolling was approximately 10% faster than the nominal
    pointer speed, that has been adjusted now.
    
    The new libinput debug-tablet tool is useful to check whether a tablet sends
    the full axis values (or any, for that matter). libinput debug-events now
    takes multiple devices, libinput record has a better CLI for recording
    multiple devices simultaneously and libinput replay had simultaneous replay
    fixed.
    
    Tablet forced proximity out has been improved - it won't happen while the
    pen is down.
    
    In libinput 1.14 we disabled a direct tool switch from tablets, causing the
    eraser button to stop working. This was caused by some tablets that break
    the spec and act as if pen and eraser are both in proximity when the button
    is pressed. This issue is fully fixed now and the eraser button should work
    as expected now.
    
    And finally, lots of fixes to the CI with the massive benefit that we can
    actually run the test suite in the CI now. Less reliance on my little laptop
    is a good thing.
    
    As usual, the git shortlog is below.
    
    Anatolii Lishchynskyi (1):
          quirks: ignore mode switch button on Madcatz R.A.T.3 mouse
    
    Björn Daase (1):
          quirks: fix horizontal scrolling for Logitech MX Master 2S on bluetooth
    
    Peter Hutterer (8):
          gitlab CI: use a minimal alpine image for the commit checks
          gitlab CI: fix skopeo copy
          gitlab CI: update to latest ci-templates
          meson.build: re-add missing trailing slash to the HTTP link
          doc/user: fix some HTML formatting
          test: add the test cases to the suite in the utils tests
          test: indentation fix
          libinput 1.15.0
    
  • Libinput 1.15 Released For Improving Input On X.Org + Wayland Desktops

    Among the changes to find with libinput 1.15 are scroll button locking, a new pad key event for special event buttons (such as for launching an on-screen display), touchpad scrolling speed fixes, a new libinput debug-tablet tool, various device fixes / quirks added, and other fixes.

  • Mesa Development Activity Was Up By ~20% In 2019, Just Under 3 Million Lines Of Code

    Mesa3D as principally the collection of Linux OpenGL/Vulkan drivers is up to 2,996,270 lines of code (and documentation / associated scripts) within its Git tree! That should put it over the three million mark very soon while the Git activity was up by about 20% in 2019.

    Mesa's nearly three million lines is spread across 7,282 files and has seen commits from over one thousand developers and amounting to 118,826 commits as of New Year's Day.

Benchmarking 9 Linux Distributions On A $50 Processor

Filed under
Graphics/Benchmarks

Your choice of Linux distribution on a budget PC can mean the difference of ~14% performance overall. Here are benchmarks of Ubuntu, CentOS, Debian, EndeavourOS, Manjaro Linux, openSUSE Tumbleweed, Fedora Workstation, and Clear Linux on a $50 processor as we roll into 2020 with the newest Linux distribution releases.

For some holiday benchmarking fun I was testing out various Linux distributions on the AMD Athlon 3000G, the recent $50 processor that features two cores / four threads, 3.5GHz clock speed, and Vega 3 graphics while having a 35 Watt TDP. Of many Linux distributions tried, the only modern Linux distribution where I ran into troubles was with Debian 10.2 stable. Even with the proprietary microcode loaded, the Vega 3 graphics weren't working with the default driver stack shipped by Debian 10.2. But aside from that it was a smooth experience on all other major distributions, including Debian Testing.

Read more

PineBook Benchmarks For The ARM Linux Laptop Starting At $99 USD

Filed under
Graphics/Benchmarks

For those interested in benchmarks of the $99+ PineBook ARM Linux laptop, more results continue to be uploaded on OpenBenchmarking.org.

The PineBook as a refresher is a 11.6" or 14" laptop with quad-core ARM Cortex-A53 processor, 2GB of RAM, 16GB eMMC, 802.11n WiFi + Bluetooth 4.0, USB 2.0 ports, microSD, a full-size keyboard, and a basic 10000mAH battery. For the price of the assembled laptop at just $99 USD isn't bad at all especially with reportedly nice build quality.

Read more

Sway 1.3 Release Candidate Brings Many Changes For This i3-Inspired Wayland Compositor

Filed under
Graphics/Benchmarks

Sway 1.3 is coming with virtual pointer protocol support needed for VNC support, repaint scheduling, layer-shell layer changes, various input improvements, support for the tablet protocol, and a variety of other fixes and enhancements.

Read more

Kernel: Linux in the 2010s, Some Benchmarks and Experimental GCN 1.0 GPU Support Might Be Dropped From AMDGPU Linux Driver

Filed under
Graphics/Benchmarks
Linux
  • The Linux Kernel Highlights Of The 2010s From Torvalds' Sabbatical To Dealing With Vulnerabilities

    Going along with our other end of year and decade recaps, here is a look back at the Linux kernel highlights for the 2010s.

    The Linux kernel during the 2010s saw a lot of new features and expanded hardware support, fallout from many security vulnerabilities and having to provide various CPU mitigations as well, Microsoft beginning to contribute to the Linux kernel largely in the context of Hyper-V, various performance improvements, debates over the state of 32-bit's future, and much more.

  • Power Management Improvements Could Benefit Intel Server Performance In Linux 5.6

    Some Intel server platforms could see better performance with the Linux 5.6 kernel cycle.

    Intel's Rafael Wysocki who also serves as the Linux kernel's power management subsystem maintainer has been queuing some patches recently in working on ACPI _CST support around the Intel-Idle driver.

  • The AppArmor Performance Impact In 70+ Benchmarks On Linux 5.5 Git

    With bisecting one of the big regressions in Linux 5.5 and finding the culprit to be an AppArmor change while using Hackbench as one of the most affected tests, I was curious to see what other workloads are impacted big by AppArmor on the current Linux 5.5 Git code. Here are 72 tests with the Threadripper 3970X on Linux 5.5 Git when toggling AppArmor.

    These New Year's Eve benchmarks are looking at the performance of Linux 5.5 Git as of two days ago when running out-of-the-box on Ubuntu 19.10 and then booting with apparmor=0 to force AppArmor to be disabled. Thus looking at the overall cost of AppArmor on Linux 5.5 right now as opposed to just the change from the recent regression.

  • The Experimental GCN 1.0 GPU Support Might Be Dropped From AMDGPU Linux Driver

    By default the Linux kernel selects the aging Radeon DRM driver for GCN 1.0 "Southern Islands" and GCN 1.1 "Sea Islands" hardware (as well as all older ATI/AMD GPUs) while it's GCN 1.2 and newer that defaults to the modern AMDGPU kernel driver. But for years there has been experimental GCN 1.0/1.1 support available via kernel module options, but now for the original GCN GPUs that code is at risk of being dropped.

Syndicate content