Language Selection

English French German Italian Portuguese Spanish

Graphics/Benchmarks

Kernel/Graphics: AMD, Intel and Mesa

Filed under
Graphics/Benchmarks
Linux
  • AMD vs. Intel Contributions To The Linux Kernel Over The Past Decade

    Driven by curiosity sake, here is a look at how the total number of AMD and Intel developers contributed to the upstream Linux kernel during the 2010s as well as the total number of commits each year from the respective hardware vendors. 

    These numbers were obtained by looking at the Linux kernel commits in Git from AMD.com and Intel.com addresses. Granted, sometimes developers from both companies will use their personal email addresses rather than the corporate ones, but for this comparison is looking solely at the Git commits from the respective corporate domains.

  • Linux k10temp Driver For AMD CPUs Updated To Better Handle Power/Temp Analysis

    As we have been eagerly talking about for the past week, the Linux kernel's k10temp driver was updated for better AMD CPU CCD temperatures and voltage/current reporting. Those improvements have been quickly evolving thanks to the work of the open-source community with AMD still sadly holding the datasheets concerning the power/temperature registers close to their vest. A new version of k10temp was sent out on Wednesday. 

    As reported earlier this week, these k10temp improvements could land for the upcoming Linux 5.6 but additional testing is needed. While Zen 2 CPUs have been shipping for months, these k10temp improvements are only coming now thanks to HWMON maintainer Guenter Roeck who continues working on this driver in cooperation with the community as AMD currently isn't releasing documentation/datasheets concerning the power/thermal registers or any reference code for that matter... Many Linux desktop users dream of seeing something someday like AMD Ryzen Master coming to Linux. 

  • Gutting Out Intel MPX Support To Be Finished Up In The Linux 5.6 Kernel

    The Linux support for Intel MPX has already been pretty much dead since the GCC 9 compiler dropped support for MPX. Kernel developers following that began working to remove MPX from the kernel over not having the compiler support, MPX not being widely used, and also not much code movement on the kernel side. Memory Protection Extensions (MPX) was talked up years ago by Intel for allowing the checking of pointer references at run-time to avoid buffer overflows and other potential related vulnerabilities. But in reality it didn't become too popular with developers while AddressSanitizer and other compiler sanitizer infrastructure has become more used and without the need for special bits in the CPU. Intel themselves meanwhile have deprecated MPX and say the support won't be available on future CPUs, hence not being concerned much about the Linux support departing.

  • Mesa 20.0 branchpoint planned for 2020/01/29, Milestone opened
    Hi list, due to some last minute changes in plan I'll be managing the 20.0
    release. The release calendar has been updated, but the gitlab milestone wasn't
    opened. That has been corrected, and is here
    https://gitlab.freedesktop.org/mesa/mesa/-/milestones/9, please add any issues
    or MRs you would like to land before the branchpoint to the milestone.
    
    Thanks,
    Dylan
    
  • Mesa 20.0 Feature Development Is Ending Next Week

    Mesa developers are planning to end feature work on Mesa 20.0 next week as this first quarter update to the Mesa 3D graphics stack.

    There has been a heck of lot building up for Mesa 20.0 including many ACO optimizations, many RadeonSI and RADV improvements around GFX10/Navi, Intel Gallium3D improvements, OpenGL 4.6 with NIR by default for RadeonSI, NIR support for LLVMpipe, Vulkan 1.2 for Intel ANV and Radeon RADV, and a whole lot more... My usual feature overview will be out after the code has been branched.

Graphics: Dav1d AV1 Acceleration, AMDVLK and Sway 1.4

Filed under
Graphics/Benchmarks
  • Dav1d AV1 Decoder Begins Adding AVX-512 Optimizations For Intel Ice Lake

    Ahead of the forthcoming dav1d 0.6 release, this open-source AV1 video decoder has begun implementing AVX-512 optimizations targeting Intel Ice Lake processors.

    The work has begun on AVX-512 optimizations focused on Ice Lake for this already quite speedy AV1 video decoder.

  • AMDVLK 2020.Q1.1 Brings Some Performance Tuning, Still On Vulkan 1.1

    Out this morning is AMDVLK 2020.Q1.1 as AMD's first official open-source Vulkan driver code drop of the new year.

    While the Radeon Software Adrenalin Edition driver for Windows was recently updated with Vulkan 1.2 support, this AMDVLK release is still on Vulkan 1.1 but at least updated against API 1.1.130 compliance. Hopefully their next code drop will have the Vulkan 1.2 support officially exposed. Meanwhile Mesa's RADV Radeon Vulkan driver has been supporting Vulkan 1.2 since hours after the specification's unveil.

  • Sway 1.4 Wayland Compositor Brings VNC Support, Initial Bits For MATE Panel Support

    Sway 1.4 is out today as the newest version of this i3-inspired Wayland compositor that has a growing following.

    Sway 1.4 consists of nearly 200 changes from over 50 contributors, showing the significant progress of this Wayland compositor that has been quick to pick-up features over the past few years.

Debian 7 Through Debian Testing Benchmarks With/Without Mitigations

Filed under
Graphics/Benchmarks

As part of our many Linux benchmarks in ending out the 2010s we ran tests looking at CentOS 6 through CentOS 8, seven years of Ubuntu Linux performance, and various other Linux distribution benchmarks and testing other important pieces of open-source software over time. One of the additional comparisons now wrapped up is looking at the performance of Debian GNU/Linux going back from the old 7 series through the current 10 stable series and also Debian Testing. Tests where relevant were done out-of-the-box with the default security mitigations and again with mitigations disabled.

Read more

Graphics: Wayland 1.18 Alpha, Linux on Embedded Ryzen with Radeon, and Keith Packard's X Talk

Filed under
Graphics/Benchmarks
  • wayland 1.17.91
    This is the alpha release for Wayland 1.18. Here's a highlight of the
    biggest changes:
    
    - Add support for the Meson build system (autotools is still supported
      but will be removed in a future release)
    - Add API to tag proxy objects to allow applications and toolkits to
      share the same Wayland connection
    - Track wayland-server timers in user-space to prevent creating too
      many FDs
    - Add wl_global_remove, a new function to mitigate race conditions with
      globals
    
    Thanks to all contributors!
    
    Full commit history below.
    
    Antonio Borneo (1):
          log: remove "%m" from format strings by using strerror(errno)
    
    Daniel Stone (2):
          build/doc: Ensure destination dir exists despite VPATH
          display-test: Remove unused variables
    
    Drew DeVault (3):
          Document unusual wl_registry.bind new_id behavior
          Add .editorconfig
          Improve description of wl_surface
    
    Emmanuel Gil Peyrot (2):
          cursor: Use memfd_create() when available
          wayland-shm: Don’t set SIGBUS handlers on unshrinkable fd
    
    Emmanuele Bassi (2):
          Support running tests from different build directories
          Add Meson build
    
    Harish Krupo (2):
          docs: Abort configure if docbook-xsl package is missing
          wayland.xml: document invalid_finish error in wl_data_offer.finish
    
    Jiayuan Ren (1):
          adding O_RDWR flag in the open()
    
    Jonas Ådahl (1):
          proxy: Add API to tag proxy objects
    
    Joshua Watt (2):
          scanner: Add configure check for strndup
          Move wl_priv_signal to wayland-server-private.h
    
    Leonid Bobrov (1):
          configure: detect libdl and librt
    
    Liu Wenlong (1):
          server: Fix fake "Address already in use" error
    
    Manuel Stoeckl (13):
          scanner: error when element names will not compile
          tests: Verify that wayland_scanner can catch bad identifiers
          protocol: clarify wl_display.delete_id description
          connection: do not abort when dup(fd) fails
          client: Ignore new requests if display has a fatal error
          client: Don't abort when sending a request fails
          tests: Test that send overflow doesn't abort
          tests: Fix race condition in send overflow test
          tests: Ensure that overflow test always overflows
          event-loop-test: Verify proper timer cancellation
          event-loop-test: Confirm distant timers do not fire
          event-loop: Track timer event sources in userspace
          event-loop-test: Add test to verify timer ordering
    
    Marty E. Plummer (1):
          scanner: prepend protocol name to types symbol
    
    Michael Forney (3):
          Use wl_container_of internally
          Avoid pointer arithmetic on `void *`
          protocol: fix typo in wl_data_offer.set_actions description
    
    Mosè Giordano (1):
          Add $(RT_LIBS) to fixed-benchmark LD dependencies
    
    Pekka Paalanen (2):
          configure.ac: reopen master for regular development
          scanner: include config.h from command line
    
    Scott Anderson (1):
          wayland.xml: Make releases for multiple 'wl_surface.attach' undefined
    
    Simon Ser (22):
          Add releasing.txt
          releasing: adapt for Wayland
          releasing: fixup section numbers
          protocol: allow to send a zero output refresh rate
          client: check event opcode in queue_event
          Update .editorconfig for Python
          Add an automated script to update wl_shm.format
          protocol: add a comment about the wl_shm.format script
          protocol: sync wl_shm.format with libdrm 2.4.99
          server: check global interface on bind
          tests: test that binding to a global with an interface mismatch fails
          protocol: invalid_method is sent on malformed request
          server: add wl_global_set_user_data
          server: add wl_global_remove
          tests: add a test for wl_global_remove
          build: check wayland-scanner version
          Revert "build: check wayland-scanner version"
          meson: use strict wayland-scanner mode
          autotools: use strict wayland-scanner mode
          build: check wayland-scanner version
          protocol: add missing enums for wl_data_device_manager.dnd_action
          build: bump to version 1.17.91 for the alpha release
    
    asynts (1):
          doc: Expand the abbreviation "hw" to "hardware".
    
    orbea (1):
          Add a missing -pthread to fix compile with slibtool.
    
    git tag: 1.17.91
    
  • Wayland 1.18 Alpha Released With Meson Support, Connection Sharing

    Wayland 1.18 is adding Meson build system support so that Autotools can be dropped in a future release, API support for allowing applications and toolkits to share the same Wayland connection, better handling over file descriptors, and wl_global_remove as a new function for mitigating race conditions with globals. There are also various test improvements, improved documentation, and various other fixes and minor improvements.

  • Linux on Embedded Ryzen with Radeon

    American Micro Devices (AMD) has released the Ryzen processors which works very well with Linux. The embedded processor also contains a Graphical Processing Unit (GPU) which is an AMD Ryzen Vega.

    The board I am using is the IBase 918f-1605 to install Linux. Linux can be installed from a stock ISO, but the system does not perform as well unless using a special Linux kernel from AMD. It also helps to have the proper GPU driver for performance. Stability is much better with the AMD kernel they provide on their website.

  • Keith Packard Talks About The Early Politics Of X Window System + Code Licensing

    At last week's Linux.Conf.Au conference was an interesting presentation by longtime X developer Keith Packard on the early days of the pre-X.Org X Window System, the collapse of Unix, and how his views formed on copyleft licenses for building thriving communities.

    Keith's LCA 2020 presentation is focused on the X happenings largely during the 80's and very early 90's. Keith's involvement goes back to the 80's during which he was employed at MIT as part of the X Consortium.

  • Keith Packard: lca2020

    I just got back from linux.conf.au 2020 on Saturday and am still adjusting to being home again. I had the opportunity to give three presentations during the conference and wanted to provide links to the slides and videos.

AMD Radeon RX 5600 XT Linux Gaming Performance

Filed under
Graphics/Benchmarks
Gaming

As announced back at CES, the Radeon RX 5600 XT is being launched as the newest Navi graphics card to fill the void between the original RX 5700 series and the budget RX 5500 XT. The Radeon RX 5600 XT graphics cards are beginning to ship today at $279+ USD price point and offers great Linux support but with one last minute -- and hopefully very temporary -- caveat.

The Radeon RX 5600 XT features 36 compute units, 2304 stream processors, up to 7.19 TFLOPs, a 1375MHz game clock, 6GB of GDDR6 video memory, and a total board power of around 150 Watts. The Radeon RX 5600 XT like the rest of the RDNA/Navi line-up is a 7nm part, supports PCI Express 4.0, and other common RDNA features.

Read more

XanMod-ing Ubuntu To Perform Closer To Intel's Clear Linux

Filed under
Graphics/Benchmarks

Earlier this month many Phoronix readers were interested in our fresh tests of the XanMod-patched Linux kernel for boosting the desktop and workstation performance compared to Ubuntu's default Linux kernel. Among many patches, XanMod does pull in some kernel patches from Intel's performance-optimized Clear Linux, so we figured it would be interesting to see how the XanMod'ed Ubuntu compares to Clear Linux performance.

As covered more in the earlier article, the XanMod Linux kernel flavor makes use of the BFQ I/O scheduler, offers CPU scaling governor improvements, makes use of preemptive full tickless kernel settings, and has a variety of other patches from leveraging Clear Linux optimizations to the BMQ process scheduler to the Proton FSYNC patches to much more. This round of testing was using a daily snapshot of Ubuntu 20.04 LTS with its current Linux 5.4 default kernel and then re-tested using the same Ubuntu 20.04 LTS installation but running on the 4.1.10-xanmod6 kernel at the time. Additionally, the same CFLAGS/CXXFLAGS as Clear Linux defaults to were also utilized.
Read more

The Performance Cost To SELinux On Fedora 31

Filed under
Graphics/Benchmarks

Following the recent AppArmor performance regression in Linux 5.5 (since resolved), some Phoronix readers had requested tests out of curiosity in looking at the performance impact of Fedora's decision to utilize SELinux by default. Here is how the Fedora Workstation 31 performance compares out-of-the-box with SELinux to disabling it.

By default Fedora runs with SELinux enabled in an enforcing and targeted mode. But by booting with selinux=0 as a kernel parameter or editing /etc/selinux/config it's possible to outright disable the Security Enhanced Linux functionality or change its operating mode.

Read more

AMD: Ryzen, AMDGPU and More

Filed under
Graphics/Benchmarks
Linux
Hardware
  • ASUS TUF Laptops With Ryzen Are Now Patched To Stop Overheating On Linux

    The AMD Ryzen Linux laptop experience continues improving albeit quite tardy on some elements of the support. In addition to the AMD Sensor Fusion Hub driver finally being released and current/voltage reporting for Zen CPUs on Linux, another step forward in Ryzen mobile support is a fix for ASUS TUF laptops with these processors.

  • AMD Sends In A Bunch Of Fixes For Linux 5.6 Along With Pollock Support

    After already several rounds of feature work queued in DRM-Next for Linux 5.6, AMD has submitted a final batch of feature work for this next kernel as it concerns their AMDGPU graphics driver.

    While Linux 5.6's merge window isn't opening until around the start of February, with RC6 having come, it effectively marks an end to the feature window of DRM-Next for targeting the next kernel. AMD's final pull request is mostly centered on fixes plus a few other extras and also enabling AMD Pollock display/graphics support for that forthcoming hardware.

  • The AMD Ryzen Thermal / Power Linux Reporting Improvements Working Well - V2 Up For Testing

    A few days ago I reported on AMD's "k10temp" Linux kernel driver finally seeing the ability to report CCD temperatures and CPU current/voltage readings as a big improvement to this hardware monitoring driver. The work hasn't yet been queued for inclusion into the mainline kernel, but initial testing is working well and a second revision to the patches has been sent out.

    Linux HWMON maintainer Guenter Roeck who spearheaded this work independent of AMD sent out the "v2" k10temp driver improvements on Saturday. This allows core complex tie temperature reporting for Zen 2 CPUs and allows current and voltage reporting for Ryzen CPUs. While this information has long been available to Windows users, sadly it's not been the case for Linux at least as far as mainline drivers go -- the out-of-tree Zenpower driver and other third-party attempts have been available but nothing mainline.

Ryzen 9 3900X vs. Ryzen 9 3950X vs. Core i9 9900KS In Nearly 150 Benchmarks

Filed under
Graphics/Benchmarks

This week our AMD Ryzen 9 3950X review sample finally arrived and so we've begun putting it through the paces of many different benchmarks. The first of these Linux tests with the Ryzen 9 3950X is looking at the performance up against the Ryzen 9 3900X and Intel Core i9 9900KS in 149 different tests.

The Ryzen 9 3950X is AMD's new $749 USD processor just below the Threadripper 3960X. The 3950X offers sixteen cores / 32 threads with a 3.5GHz base clock and 4.7GHz maximum turbo clock. Over the 3900X at $499 is four extra cores / eight threads, a 300MHz lowering of the base clock, 100MHz higher maximum boost clock, and larger L1 and L2 caches while both of these Zen 2 processors share the same TDP of 105 Watts.

Read more

Also: Ryzen CPUs On Linux Finally See CCD Temperatures, Current + Voltage Reporting

Intel graphics patch "wrecks" Gen7 iGPU Linux performance

Filed under
Graphics/Benchmarks
  • Intel graphics patch "wrecks" Gen7 iGPU Linux performance

    Earlier this week Intel released details about a vulnerability in its integrated graphics hardware. Its advisory ID was INTEL-SA-00314 and it talked about the CVE-2019-14615 vulnerability. Products from 3rd Gen Core up to 10th Gen are affected including the contemporaneous Xeon, Pentium, Celeron and Atom products. Intel was made aware of this vulnerability as far back as August so already has patches available and links to recommended new drivers for both Windows and Linux users (scroll down this page about half way).

    All so regular and nothing surprising so far… However, since the updated drivers have been released, Linux-centric tech site Phoronix has been busy checking and testing the new drivers (on Linux of course) to see if there are any performance penalties, or other aberrations, delivered with the vulnerability patches.

    Intel describes the CVE-2019-14615 vulnerability as follows: "Insufficient control flow in certain data structures for some Intel Processors with Intel Processor Graphics may allow an unauthenticated user to potentially enable information disclosure via local access." Please note the key phrase - local access - but Phoronix thinks that WebGL within web browsers is another possible attack vector.

    In its Linux testing, Phoronix was initially unperturbed by results on processors sportin

  • Intel's Mitigation For CVE-2019-14615 Graphics Vulnerability Obliterates Gen7 iGPU Performance

    Yesterday we noted that the Linux kernel picked up a patch mitigating an Intel Gen9 graphics vulnerability. It didn't sound too bad at first but then seeing Ivy Bridge Gen7 and Haswell Gen7.5 graphics are also affected raised eyebrows especially with that requiring a much larger mitigation. Now in testing the performance impact, the current mitigation patches completely wreck the performance of Ivybridge/Haswell graphics performance.

    The vulnerability being discussed and analyzed this week is CVE-2019-14615. This CVE still hasn't been made public over 24 hours later (though there are the Intel SA-00314 details for this disclosure), but from going through kernel patches and other resources, it certainly caught our interest right away and have been benchmarking it since yesterday evening. The CVE-2019-14615 vulnerability amounts to a new information disclosure issue due to insufficient control flow in certain data structures. Local access is required for exploiting this control flow issue in the hardware, but it's not yet known/published if say WebGL within web browsers could exploit this issue. This is a hardware issue with all operating systems being affected. Our testing today, of course, is under Linux.

Syndicate content