What’s New in V3.5?

New Features

• New – DirectX 12 support has improved with greater stability and improved performance
• NEW – Support for LiquidVR 1.0.3.20 – Find out more about LiquidVR
• NEW – The open source version of the DirectX 12 server plugin is now supported as part of CodeXL at GPUOpen

Changes

• GPU Profiling is now only supported on GCN hardware
• This release improves stability with DirectX 12 applications
• The DirectX 12 GPU Trace profiler has been rewritten for improved performance with multi-threaded games
• FIX for client UI issues with Windows 10 Chinese Edition

Known Issues

• GPU Profiling is now only supported on GCN hardware
• Shader Analysis of GLSL shaders is only supported on 16.5.2 drivers (and earlier).

GPU PerfStudio V3.4

New Features

• NEW – Support for an open source DirectX® 12 Plugin
• With the GPUOpen initiative AMD is now providing  open source access and build environment for the GPU PerfStudio DirectX® 12 plugin. The plugin can be used to replace the version shipped with the public release of GPU PerfStudio 3.4. More information can be found at GPUOpen.

Changes

• This release improves stability with DX11 applications.
• Connection dialog change – a second dialog of settings is now presented to the user while starting an application with the client.
• Support for ID3D11Device2 and ID3D11Device3
• Support for ID3D11DeviceContext2 and ID3D11DeviceContext3
• Support for IDXGISwapChain_Present1.

Bug Fixes

• Fixed a bug with UpdateSubResource() and UpdateSubResource1() where not enough data was being stored for a buffer.
• Fix a bug with failing Maps() during frame capture playback.
• Fixed a bug in ClearRenderTargetView, ClearUnorderedAccessViewUint, ClearUnorderedAccessViewFloat, and ClearDepthStencilView which would cause the API Trace to crash in cases where the application had released the view.
• Fix for crash in the API Trace where PerfMarkers with special characters were not being escaped for XML.
• Fixed a bug when cloning a resource.

GPU PerfStudio V3.3

New Features

• NEW – Support for DirectX® 12 on Windows 10 (more information here)
  

• NEW – Support for  OpenGL 4.5 

• NEW – Support for OpenGL ES 3.1

• NEW – Support for LiquidVR™

Changes

• Connection dialog changes – added a text entry field for GPU PerfServer options
• Constant buffer changes – constant buffer names are now displayed
• Improvements to the wireline geometry rendering and render target overlay
• Changes to the server option names
• Shader Analyzer – support for 64bit compilation
• Help file has been re-ordered and improved
• API Trace files can now be saved and loaded from disk
• The HUD settings are now saved in the client settings file

Bug Fixes

• Improvements to UAV support
• Improved title support
• Improvements in previous versions

GPU PerfStudio V3.2

• NEW – OpenGL ES 3.0 Support
  • Support for OpenGL ES 3.0 on Windows and Linux

• NEW – Frame Based Instant Capture
  • The GPU PerfStudio server now has the ability to perform a frame capture at a specified frame number without client intervention
  • This feature can be used to automatically capture  a specific frame within a game or benchmark using a script or batch file

• NEW – Shader Analyzer wave occupancy view
  • This view displays the number of waves in flight based on the number of general purpose registers used by a HLSL shader
  • This feature can be used to optimize a HLSL shader based on its wave occupancy information

Bug Fixes

• When profiling with GPUTime selected, the percentage results will be weighted using the GPUTime for each draw call. Otherwise, the result will be a non-weighted average.
• GPU PerfStudio now supports the use of 64bit games when launched from 32bit launchers
• Improved title support

GPU PerfStudio V3.1

New Features

• NEW – GPU ShaderAnalyzer for GCN

Optimize your shader code using the latest version of this tool by viewing compiler statistics, ISA and D3D ASM code, and customizing your compiler settings for the latest AMD GCN GPUs.

• Profiler support for D3D11_QUERY
• Additional counters have been added to show the time spent in each stage of the GPU:
  • VSTime – Time vertex shader is busy in milliseconds
  • HSTime – Time hull shaders is in milliseconds
  • DSTime – Time domain shader is busy in milliseconds
  • GSTime – Time geometry shader is busy in milliseconds
  • PSTime – Time pixel shader is busy in milliseconds
  • CSTime – Time compute shader is busy in milliseconds
• Support for OpenGL 4.4 (Windows and Linux)

Bug Fixes

• Fix for compute shader counters showing non zero values when no compute shader is in use
• Fix for incorrect CSVALUInsts counter values
• Many stability enhancements for titles running on DirectX 11 and OpenGL
• Improved support for viewing MSAA depth buffers and render targets
• Fixes for client and server side memory leaks
• Fix for black texture images in some Steam for Linux games
• Improvements in previous versions

GPU PerfStudio V3.0

• Adds Linux Server support
• Adds OpenGL 4.3 Support
• Calls between glBegin & glEnd are now collapsed to a single entry.
• Adds support for separate shader objects in GL
• Many stability enhancements for titles running on DirectX 11 and OpenGL

• Fix for missing call data in FrameDebugger
• Fix for missing HUD data in GL applications

GPU PerfStudio V2.15 improvements

• Windows 8.1 support – fixes the issue where the client would not connect to games running on Windows 8.1
• Support for DX10 and DX10.1 has been removed
• Improved support for OpenGL applications
• Fixes for missing textures in OpenGL applications
• Fixes for missing data fields in the DX11 view descriptions
• Profiler stability has been improved
• The draw call number is now displayed in the API Trace
• Wireframe geometry overlay color can now be defined by the user

GPU PerfStudio V2.14 Improvements

• Profiler support for Hawaii GPUs
• Pipeline stage specific counters for OpenGL are now supported
• Improved stability for multi-threaded apps

GPU PerfStudio V2.13 Improvements

• New feature “PerfMarker View”. A new tree control has been added to the Frame Debugger for easy viewing and navigation of Direct3D performance markers (D3DPERF_BeginEvent and D3DPERF_EndEvent)
  • It displays perfmarker blocks as nested tree items making it easy to navigate the frame based on the structure of your perfmarkers
  • The individual draw calls within perfmarker blocks are shown by default and can be turned off if required
  • The GPUTime for perfmarker blocks and individual draw calls can be displayed
  • The rows can be ordered by draw call index to allow the user to navigate by draw call index
  • The rows can also be ordered by GPU Time to identify the most expensive perfmarker blocks and draw calls
  • The perfmarkers can be expanded and collapsed by using the right click context menu
  • Perfmarker blocks and individual draw calls can be re-profiled on demand
• Profiler support for latest AMD GPUs
• Fix for DX10 profiler counters on Southern Islands GPUs
• Option to disable the synching of the Frame Debugger, API Trace, and Profiler results windows

GPU PerfStudio V2.12 Improvements

• Improved process targeting – the user can specifiy the maximum number of process hops
• The server now responds to commands up to 33% faster than before
• Profiler counters are now more accurate for multipass profiling
• Fix for multisampled depth stencil appearing black in the client
• Fix for multisampled textures in DX10 apps not appearing in the client
• Fix for OpenGL GL_TRUE format image appearing black in the client
• Improved memory usage with map/unmap

GPU PerfStudio V2.11 Improvements

• OpenGL users with 7000 series GPU’s please use 12.8 Catalyst Display Drivers
• Support for GLSL in-game shader editing and re-insertion
• Direct3D performance marker Profiling
  • This feature allows the user to profile the performance marker items for GPU Time from the API Trace window
  • The profile shows the absolute GPU Time as well as a relative percentage GPU time w.r.t the frame GPU Time
• Expand/Collapse all Direct3D performance marker tree items in API Trace view window
• API Trace tooltip info now shows GPU Time values for Direct3D performance markers
• Add support for additional OpenGL draw call APIs , double precision formats, and sampler objects
  • glDrawArraysInstancedBaseInstance
  • glDrawElementsInstancedBaseInstance
  • glDrawElementsInstancedBaseVertexBaseInstance
  • glMultiDrawArraysIndirectAMD
  • glMultiDrawElementsIndirectAMD
  • glCopyBufferSubData
• The OpenGL pipeline stages now have the correct GL names
• Support for OpenGL GL_TEXTURE_SRGB_DECODE_EXT and sampler objects
• A new column has been added to the API Trace view window to show the CPU time spent by each API call in the driver
  • Improved GL Profile support, for faster profiles.
  • Improved support for PRO, LE and XT flavors of Juniper cards
  • Optimizations for faster DirectX11 ™ profiles on pre-7000 cards
• Support for DX11 RenderTargetBlendDesc and Stencil Ref value
• The GPU PerfStudio client now requires Microsoft .NET Framework 4

 GPU PerfStudio V2.10 Improvements:

• Support for editing of pre-compiled debug shaders
• Enhanced profiler support for Southern Islands GPUs
• Index and vertex buffer scan be saved to disk
• Texture viewer now has a Y-flip option
• GPU PerfStudio is now OpenGL Core Profile compliant
• Support for more game titles
• DX11 depth stencil states and render target blend states now display correctly
• General stability improvements in the OpenGL server
• Improved memory usage in the OpenGL profiler