Language Selection

English French German Italian Portuguese Spanish

Graphics/Benchmarks

Graphics: Mir, Zink and DXVK

Filed under
Graphics/Benchmarks
  • Ubuntu's Mir Working On Replaceable Renderer, Hybrid Graphics Driver Support

    Canonical's Chris Halse Rogers has shared a road-map for Mir (or terrain map as he prefers calling it) about their future plans for this open-source display server that remains focused now on providing Wayland support. 

  • Zink Merged Into Mesa 19.3 For Offering OpenGL Over Vulkan

    Zink is the effort led by Collabora's Erik Faye-Lund for offering a generic OpenGL/GLES implementation that runs atop Vulkan. While it's exciting prospects and well into the future could allow hardware vendors to avoid having to maintain OpenGL drivers with instead focusing on Vulkan, for now there is still a long road ahead for performance and features. Right now Zink supports just OpenGL 2.1 / OpenGL ES 2.0. With time though there are plans for supporting OpenGL 3.x/4.x and OpenGL ES 3.x functionality. At least with Zink, the existing OpenGL code inside Mesa/Gallium3D is doing much of the heavy lifting. 

  • D9VK 0.30 Released With Performance Improvements, Other D3D9 Features Now Supported

    Building off yesterday's DXVK 1.4.4 release, D9VK 0.30 is out as the similar project that implements the Direct3D 9 API atop Vulkan.

    D9VK 0.30 re-bases its code atop DXCVK 1.4.4 and has performance improvements via locking changes, avoiding the throwing out of D3DUSAGE_DYNAMIC buffers, supporting discard on non-dynamic resources, other locking changes, and other work.

Apple macOS 10.15 vs. Windows 10 vs. Ubuntu 19.10 Performance Benchmarks

Filed under
Graphics/Benchmarks

In addition to this month bringing the release of the Ubuntu 19.10 "Eoan Ermine", Apple also shipped macOS 10.15 "Catalina" as the sixteenth major release of their macOS operating system. So with that it makes for an interesting time seeing how macOS 10.15 competes against both Ubuntu 19.10 and Windows 10 on an Apple MacBook Pro. Here are those results from dozens of benchmarks.

Using an Apple MacBook Pro with Core i7-6700HQ Skylake CPU, 2 x 8GB RAM, 250GB Apple SSD, and Radeon Pro 450 graphics, macOS 10.15, Windows 10, and Ubuntu 19.10 were all benchmarked off this same system. All three operating systems were tested with their latest software updates as of testing.

Read more

DXVK 1.4.4 Released

Filed under
Graphics/Benchmarks
Software
Gaming
  • D3D11 and D3D10 to Vulkan layer DXVK version 1.4.4 is out

    One missed from the weekend, developer Philip Rebohle released a fresh update to the Vulkan layer DXVK with version 1.4.4.

    A small and sweet maintenance release we have here with a couple of optimizations. The memory footprint of small and frequently updated buffers was reduced, there's some minor optimizations for "Stream Output and Append/Consume buffers (used e.g. by Unity Engine)", a bug fixed from DXVK 1.4.3 that caused some invalid state cache entries being generated, some Vulkan validation errors with geometry shaders were fixed and some "potential read-after-write hazards involving vertex and index buffers" were also solved.

  • DXVK 1.4.4 With Vulkan Usage Fixes, Optimizations & A Few Game Specific Fixes

    Philip Rebohle has released his latest weekly update to DXVK for accelerating Direct3D 10/11 games using Vulkan as a big boost for Steam Play (Proton) and Wine.

    DXVK 1.4.4 has a regression fix for 1.4.3 that could lead to invalid Vulkan API usage, Vulkan validation error fixes, potential read-after-write hazards resolved, optimizations for Stream Output and Append/Consume buffers, and reduced memory footprint for small and frequently updated buffers.

EPYC 7642/7742 vs. Xeon Platinun 8280 Performance With Intel-Recommended Benchmarks

Filed under
Graphics/Benchmarks

Here are the latest numbers in our ongoing AMD EPYC 7002 "Rome" series benchmarking. This time around for some curiosity over the weekend is a look at the EPYC 7642 and EPYC 7742 up against the Xeon Platinum 8280 all in 2P configurations while running benchmarks publicly recommended by Intel in one of their whitepapers. Even going by Intel-recommended tests, the EPYC 7642/7742 deliver significantly better performance and cost savings over the comparable Xeon Platinum 8280.

One of my weekend "hobbies" is always scouring whitepapers, scientific papers, and GitHub (among other sources) looking for interesting and new benchmarks/workloads to incorporate into the Phoronix Test Suite / OpenBenchmarking.org. After all, there are 384 different official benchmarks part of our open-source benchmarking framework and from there 1,404 different versions of those benchmarks from the past 11+ years of developing the Phoronix Test Suite for open-source and fully-automated benchmarking.

Read more

Graphics: AMD, Vulkan and NVIDIA

Filed under
Graphics/Benchmarks
  • AMD Sends In Another Batch Of Feature Changes For Linux 5.5 AMDGPU Driver Code

    Complementing the big AMDGPU feature pull request from two weeks ago, on Friday AMD sent out a second batch of features targeting the upcoming Linux 5.5 kernel merge window.

    On the feature front, the main change with this pull request for Linux 5.5 is the BACO support for older AMD GPUs. Those Radeon 200/300 series parts are seeing BACO support for possible power-savings ultimately but the short-term focus is using it for GPU reset functionality. BACO is already supported on the newer AMD GPUs.

  • RADV Vulkan Driver's ACO Compiler Back-End Has Navi/GFX10 Nearly Squared Away

    Following a lot of work on the AMD "ACO" compiler back-end to the RADV Vulkan driver for GFX10/Navi, this experimental alternative to AMDGPU LLVM is about ready to go for these newest AMD graphics processors.

    There was more aco/gfx10 activity merged yesterday getting the hazard mitigations in place and other bits landing earlier in the week as well.

  • NVIDIA Extends GL_SYNC_DISPLAY_DEVICE To Work With Vulkan Apps, Better RAM Reclamation

    NVIDIA on Friday released the 435.27.02 Linux beta driver that features a few Vulkan updates.

    First up, the environment variable __GL_SYNC_DISPLAY_DEVICE is now honored by Vulkan applications. This environment variable has been used by the NVIDIA Linux driver for a few years in cases of multi-monitor setups that are not ideal for specifying which display device / output to sync to in a bid to eliminate tearing. This GL_SYNC variable has obviously been focused on OpenGL (there is also a separate one for VDPAU synchronization) while now it works for Vulkan games/programs too.

A Look At The Per-Clock Performance / Peak Frequencies With The Intel Core i7-1065G7 Ice Lake

Filed under
Graphics/Benchmarks

Following this week's Intel Core i7-1065G7 Ice Lake Linux benchmarks there was some questions and speculations about the per-clock performance of this long-awaited Intel microarchitecture update. Here is some additional data shedding light on the clock frequencies under load and ultimately how the per-clock performance compares to the common Intel previous-generation mobile CPUs.

The Core i7-1065G7 features a 1.3GHz base frequency and 3.9GHz turbo frequency, well below the i7-8550U clock frequencies of 1.80GHz and 4.0GHz for the i7-8550U or 1.8GHz and 1.6GHz of the i7-8565U processors, the two other mobile processors compared to in this week's Linux benchmarks. But thanks to superior per-clock performance, the Ice Lake CPU offered significant raw performance and performance-per-Watt gains over these earlier processors.

Read more

Linux Graphics: Intel's ANV Driver, Mesa 19.2.2, NVIDIA Vulkan Beta Driver 435.27.02

Filed under
Graphics/Benchmarks
Linux
  • Intel's ANV Driver Now Exposes Vulkan Memory Model Support

    The newest Vulkan extension now supported by Intel's open-source "ANV" Linux driver is VK_KHR_vulkan_memory_model.

    The VK_KHR_vulkan_memory_model support was originally introduced a year ago and then solidified in Q1 as the formal memory model for Vulkan and the first for a major graphics API. VK_KHR_vulkan_memory_model exposes support for the Vulkan Memory Model as the means of defining synchronized memory access to the same memory locations via multiple shader invocations.

  • mesa 19.2.2
    Hi list,
    
    Sorry for the slight delay in the release, my fault. I'm pleased to announce the
    availability of mesa 19.2.2, the second point release in the mesa 19.2 series.
    Things have started to slow down a bit, and I like that. We've had an assortment
    of fixes in this release, notably a bunch of work to get Solaris and illumos
    working with mesa, as well as more work fixing issues in the migration of
    package-config and headers being handled by libglvnd instead of mesa when mesa
    is built with support for glvnd.
    
    There's  bunch of other changes here, with radv and intel leading the pack,
    otherwise just a few things here and there.
    
    Dylan
    
    
    Shortlog
    ========
    
    Alan Coopersmith (6):
          c99_compat.h: Don't try to use 'restrict' in C++ code
          util: Make Solaris implemention of p_atomic_add work with gcc
          util: Workaround lack of flock on Solaris
          util: Solaris has linux-style pthread_setname_np
          meson: recognize "sunos" as the system name for Solaris
          intel/common: include unistd.h for ioctl() prototype on Solaris
    
    Alejandro Piñeiro (1):
          v3d: take into account prim_counts_offset
    
    Bas Nieuwenhuizen (3):
          radv: Disallow sparse shared images.
          nir/dead_cf: Remove dead control flow after infinite loops.
          radv: Fix single stage constant flush with merged shaders.
    
    Clément Guérin (1):
          radeonsi: enable zerovram for Rocket League
    
    Connor Abbott (2):
          nir/sink: Rewrite loop handling logic
          nir/sink: Don't sink load_ubo to outside of its defining loop
    
    Dylan Baker (3):
          docs: Add SHA256 sum for 19.2.1
          docs: Add release notes for 19.2.2
          Bump version for 19.2.2 release
    
    Eric Engestrom (7):
          GL: drop symbols mangling support
          meson: rename `glvnd_missing_pc_files` to `not glvnd_has_headers_and_pc_files`
          meson: move a couple of include installs around
          meson: split headers one per line
          meson: split Mesa headers as a separate installation
          meson: skip installation of GLVND-provided headers
          util/u_atomic: fix return type of p_atomic_{inc,dec}_return() and p_atomic_{cmp,}xchg()
    
    Ian Romanick (2):
          nir/search: Fix possible NULL dereference in is_fsign
          intel/vec4: Don't try both sources as immediates for DPH
    
    James Xiong (1):
          iris: finish aux import on get_param
    
    Kenneth Graunke (2):
          iris: Properly unreference extra VBOs for draw parameters
          iris: Implement the Gen < 9 tessellation quads workaround
    
    Lepton Wu (1):
          egl/android: Remove our own reference to buffers.
    
    Lionel Landwerlin (3):
          etnaviv: remove variable from global namespace
          anv: fix vkUpdateDescriptorSets with inline uniform blocks
          anv: fix memory leak on device destroy
    
    Lucas Stach (3):
          etnaviv: fix vertex buffer state emission for single stream GPUs
          rbug: fix transmitted texture sizes
          rbug: unwrap index buffer resource
    
    Pierre-Eric Pelloux-Prayer (1):
          mesa: fix invalid target error handling for teximage
    
    Roland Scheidegger (1):
          gallivm: Fix saturated signed psub/padd intrinsics on llvm 8
    
    Samuel Pitoiset (6):
          drirc: enable vk_x11_override_min_image_count for DOOM
          radv: bump minTexelBufferOffsetAlignment to 4
          radv: fix DCC fast clear code for intensity formats
          Revert "radv: do not emit PKT3_CONTEXT_CONTROL with AMDGPU 3.6.0+"
          radv: fix DCC fast clear code for intensity formats (correctly)
          radv: fix updating bound fast ds clear values with different aspects
    
    Timothy Arceri (1):
          glsl: fix crash compiling bindless samplers inside unnamed UBOs
    
    
    git tag: mesa-19.2.2
    
  • Mesa 19.2.2 Brings Fixes For Solaris/Illumos, Continued Fixes For Intel ANV + Radeon RADV

    Mesa 19.2.2 was released on Thursday as the second point release to this quarter's Mesa 19.2 stable series.

    With all new feature work focused on Mesa 19.3 for release in just over one month's time, Mesa 19.2.2 is just focused on fixes. The Mesa 19.2.2 brings a number of compatibility fixes/workarounds for Oracle Solaris (and the open-source Illumos), a number of Radeon Vulkan (RADV) fixes, the workaround to enabling zeroed vRAM for the Rocket League game to avoid artifact issues, Meson build system updates, a few Intel Iris Gallium3D driver fixes, and various other fixes scattered throughout.

  • NVIDIA have released a new Vulkan Beta Driver 435.27.02

    NVIDIA continue pushing out new builds to their special Vulkan Beta Driver, a staging area to test out new features. They've been running this special series now for a number of years, as they continue to keep up with the latest updates to the Vulkan API and their support of Linux with recent drivers has been great.

Ubuntu 19.04 vs. 19.10 Performance On High-End AMD/Intel Desktop CPUs

Filed under
Graphics/Benchmarks

For those curious how the Ubuntu 19.04 vs. Ubuntu 19.10 performance is looking for higher-end Intel/AMD desktops, here are some fresh benchmarks.

With last week's release of Ubuntu 19.10 we began carrying out some Ubuntu 19.04 vs. 19.10 performance comparisons. Based upon the higher-end desktop CPUs we had access to, this comparison is of four different systems including the Intel Core i9 9900K, Intel Core i9 7980XE, AMD Ryzen 9 3900X, and AMD Ryzen Threadripper 2990WX.

Note there are component differences between the systems with this testing is focused on looking at the Ubuntu 19.04 vs. 19.10 performance as opposed to directly comparing the various systems under test, which we do anyhow in our various Linux CPU reviews.

Read more

Zink: Fall Update

Filed under
Graphics/Benchmarks

I recently went to XDC 2019, where I gave yet another talk about Zink. I kinda forgot to write a blog-post about it, so here’s me trying to make up for it… or something like that. I’ll also go into some more recent developments as well.

My presentation was somewhat similar to the talk I did at SIGGRAPH this year, but with a bit more emphasis on the technical aspect, as the XDC audience is more familiar with Mesa internals.

If you’re interested, you can find the slides for the talk here. The talk goes through the motivation and basic approach. I don’t think I need to go through this again, as I’ve already covered that before.

As for the status, Zink currently supports OpenGL 2.1 and OpenGL ES 2.0. And there’s no immediate plans on working on OpenGL 3.0 and OpenGL ES 3.0 until Zink is upstream.

Which gets us to the more interesting bit; that I started working on upstreaming Zink. So let’s talk about that for a bit.

Read more

Intel Icelake "Gen11" Graphics Are A Huge Upgrade Over Gen9 With Good Linux Support

Filed under
Graphics/Benchmarks

Similar to the Linux CPU performance article earlier this week, the Core i7-1065G7 was tested via a Dell XPS 7390 that was purchased retail for being able to deliver these Linux benchmarks. The Icelake results were compared to the other Dell XPS models I had available as the Dell XPS 9380 with Core i7 8565U Whiskey Lake and Dell XPS 9370 with Core i7 8550U Kabylake-R for showing the generational graphics performance comparison with these otherwise similar laptops. A larger Linux laptop comparison with older laptop models and more will be coming in the next few weeks for helping Linux users better evaluate upgrade options ahead of this holiday season.

As a refresher, the Core i7 8550U and Core i7 8565U features "Gen 9" UHD Graphics 620 while the new Icelake Core i7-1065G7 features "Gen 11" Iris Plus Graphics.

Read more

Syndicate content