Recent Posts

Pages: [1] 2 3 ... 10
3D-Tech News Around The Web / Vulkan API Overhead Test Added in 3DMark
« Last post by JeGX on Today at 03:18:39 PM »
Futuremark has released a new version of 3DMark that adds the Vulkan support to the API Overhead test. We can now compare the API performance of Direct3D 11, Direct3D 12 and Vulkan.
Geeks3D's GPU Tools / GPU Shark 0.9.11.x released
« Last post by JeGX on Today at 02:51:14 PM »
GPU Shark has been released. Complete story and download:
3D-Tech News Around The Web / Microsoft PIX 1703.20.002
« Last post by Stefan on March 22, 2017, 07:17:46 PM »
Performance tuning and debugging for DirectX 12 games on Windows and Xbox
PIX 1703.20.002 – NVIDIA GPU counters, file IO traces, shader edit & continue, and pixel cost visualizer

Today we released PIX 1703.20.002 beta.  New in this release:
  • NVIDIA GPU counters
  • File IO traces
  • Shader edit & continue
  • Pixel Cost visualizer
  • Improved Warnings UI.  Where possible, warning hyperlinks now point to the GPU rather than CPU event list.
  • Added an option to disable Capture GPU timings when taking timing captures.  Turning this off can reduce measurement overhead if you only care about CPU timing information.
3D-Tech News Around The Web / Vulkan Multi-GPU Support is not limited to Windows 10
« Last post by JeGX on March 22, 2017, 01:36:33 PM »
At GDC 2017, in San Francisco during February, Khronos™ released several new Vulkan® extensions for cross-platform Virtual Reality rendering and multiple GPU access. This functionality has been initially released as KHX extensions to enable feedback from the developer community before being incorporated into final specifications.

One key question that we have been asked since GDC is whether the Vulkan multi-GPU functionality is specifically tied to ship only on Windows 10.

The good news is that the Vulkan multi-GPU specification is very definitely NOT tied to Windows 10. It is possible to implement the Vulkan multi-GPU extension on any desktop OS including Windows 7, 8.X and 10 and Linux.

Some of the Khronos GDC presentations mentioned that for Vulkan multi-GPU functionality, Windows Display Driver Model (WDDM) must be in Linked Display Adapter (LDA) mode. That was not a very clear statement that has caused some confusion. And so it is worth clarifying that:

1/ The use of WDDM is referring to the use of Vulkan multi-GPU functionality on Windows. On other OS, WDDM is not necessary to implement the Vulkan multi-GPU extension.

2/ On Windows, the use of LDA mode can make implementing Vulkan multi-GPU functionality easier, and will probably be used by most implementations, but it is not strictly necessary.

3/ If an implementation on Windows does decide to use LDA mode, it is NOT tied to Windows 10. LDA mode has been available on many versions of Windows, including Windows 7 and 8.X.

Khronos always strives to make its specifications as cross platform as possible. Of course, what products ship on which OS is up to the implementers of each specification, but Khronos is already aware of vendor plans to ship multi-GPU functionality on platforms other than Windows 10, including Linux.


English forum / Shadertoy demos in Vulkan (Geomechanical)
« Last post by JeGX on March 22, 2017, 11:35:42 AM »
Another shadertoy demo ported to Vulkan:

Will be available in the next update of the code sample pack.
3D-Tech News Around The Web / WebGLNext: Obsidian API Proposal
« Last post by JeGX on March 22, 2017, 09:36:05 AM »
This is Mozilla's draft proposal for the GPU API for the Web, called Obsidian. It is a low-level API that provides maximum feature set of the GPU to the web applications. The API is designed for WebAssembly, modern GPUs, and multi-threaded environment in mind.

Obsidian is a temporary code name, signifying the Vulkan roots of the API:

Obsidian is a naturally occurring volcanic glass formed as an extrusive igneous rock.
This proposal is not a specification. It includes reasoning for the design decisions, draft WebIDL and a bit of example code. We don't aim to provide a complete specification, instead we want this proposal to represent our vision of the future API in the working group discussions, a vision of rich graphics on the Web powered by a low-level explicit API.

The contents are split into the following sections:

1/ introduction and philosophy
2/ design details, differences from Vulkan/D3D12
3/ synchronization and memory model
4/ API and examples

Link to the complete proposal:
3D-Tech News Around The Web / Live VGPR Analysis with Radeon GPU Analyzer
« Last post by Stefan on March 21, 2017, 10:06:24 PM »
This article explains how to use Radeon GPU Analyzer (RGA) to produce a live VGPR analysis report for your shaders and kernels. Basic RGA usage knowledge is assumed.
3D-Tech News Around The Web / AMD Vega 10 Support added in Linux drivers
« Last post by JeGX on March 21, 2017, 09:12:30 PM »
This patch set adds support for vega10.  Major changes and supported
- new vbios interface
- Lots of new hw IPs
- Support for video decode using UVD
- Support for video encode using VCE
- Support for 3D via radeonsi
- Power management
- Full display support via DC
- Support for SR-IOV


Vega 10 device support:

    {0x1002, 0x6860, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_VEGA10},
    {0x1002, 0x6861, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_VEGA10},
    {0x1002, 0x6862, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_VEGA10},
    {0x1002, 0x6863, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_VEGA10},
    {0x1002, 0x6867, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_VEGA10},
    {0x1002, 0x686c, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_VEGA10},
    {0x1002, 0x687f, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_VEGA10},

Polaris 12 device support:

    {0x1002, 0x6980, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_POLARIS12},
    {0x1002, 0x6981, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_POLARIS12},
    {0x1002, 0x6985, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_POLARIS12},
    {0x1002, 0x6986, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_POLARIS12},
    {0x1002, 0x6987, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_POLARIS12},
    {0x1002, 0x6995, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_POLARIS12},
    {0x1002, 0x699F, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_POLARIS12},

Other related links:
- AMD Sends Out 100 Patches, Enabling Vega Support In AMDGPU DRM
- AMD GPU Linux driver patches lists seven Vega 10 IDs
- AMD's Vega 10 GPU has appeared in Linux drivers

3D-Tech News Around The Web / Vulkan API specifications 1.0.44 released
« Last post by Stefan on March 20, 2017, 11:09:39 PM »
Change log for March 17, 2017 Vulkan 1.0.44 spec update:

  * Bump API patch number and header version number to 44 for this update.

Github Issues:

  * Fix description of <<features-extentperimagetype, Allowed Extent Values
    Based On Image Type>> (public issue 290).
  * Better specify VK_DEVICE_LOST behavior around flink:vkQueueSubmit,
    flink:vkWaitForFences, and flink:vkGetFenceStatus (public issue 423).
  * Clarify definition of flink:vkGetQueryPoolResults::pname:queryCount
    (public issue 441).
  * Simplify and clean up normative language. Remove shall and replace
    recommend and variants with should wherever possible (public issue 448).
  * Fix all dangling internal cross-references in the 1.0-extensions
    specification, and add scripts/checkXrefs to find these in the future
    (public issue 456).
  * Reverse order of ChangeLog.txt entries so the most recent version is
    documented first (public issue 463)
  * Removes "become invalid" which clashes with invalid state for command
    buffers. (public issue 467)
  * Disallowed pending state in spec text for vkResetCommandBuffer, matching
    valid usage (public issue 468)
  * Removes sentence describing invalid state "like initial state". (public
    issue 469)
  * Disallows begin command buffer from resetting command buffers in the
    "recording" state. (public issue 470)
  * Removes mention of state from description of
  * Removed extra valid usage statement in VkSubmitInfo (public issue 472)

Internal Issues:

  * Clarify description of the pname:imageLayout member of
  * Fix typos where etext:VK_VIEW_TYPE* was used instead of
  * Removed the <<VK_KHR_display>> and <<VK_KHR_display_swapchain>> example
    code from the specification and noted it has been moved to the Vulkan
    SDK cube demo (internal issue 179).
  * Reorder VkExternalMemoryHandleTypeFlagBitsNV description (internal issue
  * Clarify than an implementation is
    <<fundamentals-validusage-flags,permitted to return 'undefined' bit
    flags>> in a bitfield (internal issue 640).
  * Break Valid Usage statements describing unrelated parameters into
    separate statements, and add a style guide entry to follow this approach
    (internal issue 685).
  * Move valid usage statement for slink:VkImageCreateInfo from spec body to
    the explicit valid usage block (internal issue 693).
  * Fix typos in the descriptions of slink:VkDisplaySurfaceCreateInfoKHR,
    flink:vkCreateDisplayModeKHR, and
    flink:vkGetDisplayPlaneSupportedDisplaysKHR in the <<display,Presenting
    Directly to Display Devices>> section (internal issue 698, 704, 716).
  * Clarified that mandatory depth/stencil formats are only a requirement
    for 2D images (internal issue 719).
  * Clarify that variables decorated with DeviceIndex/ViewIndex must be in
    the Input storage class (internal issue 733).
  * Work around generator script problem with removal of Unicode literals
    from Python 3.0-3.2 using `future` package (internal issue 737).
  * Remove nonexistent structure type enums from vk.xml (internal issue
  * Fix validextensionstructs attributes for structures in the pname:pNext
    chain for VkPresentInfoKHR, fixing implicit valid usage statements for
    those structures (internal issue 740).

New Extensions: ---

Falcor is a Direct3D 12 real-time rendering framework. It aims to improve productivity of research and prototype projects. Its features include:

- Abstract many common graphics operations, such as shader compilation, model loading and scene rendering.
- VR support using OpenVR.
- Common rendering effects such as shadows and post-processing effects.

This is an alpha version. The interfaces are not final yet and there might be some performance/stability issues.
If you'd like OpenGL support, please consider using Falcor 1.0 ('rel-1.0' branch). Note that though Falcor 1.0 is not supported anymore, it is stable and optimized.

More information and source code:
Pages: [1] 2 3 ... 10