User Guide

  • 2020.4
  • 12/10/2020
  • Public Content

Pane: Visualization

Enables you to adjust the manner of displaying events in the
bar chart and
Visualizes performance in the frame or selected events over time, if metrics are available for the target system.
Enables you to navigate along the frame and select events that are shown on the
bar chart.
Visualization Settings Toolbar
Use the
Visualization Settings
toolbar to configure the
bar chart mode:
  • Hotspot Mode
    button - enables a Hotspot analysis. This button becomes active, if the system supports the full list of Intel performance counters.
Hotspot analysis is an advanced performance analysis that discovers performance bottlenecks in individual events. This analysis type also allows you to group the events by the causes of bottlenecks and by pipeline states to highlight the most pronounced performance issues and to see the events that should be optimized in conjunction.
Clicking the
Hotspot Mode
button changes data representation in all Graphics Frame Analyzer components. The bar charts start showing grouped calls, with the longest group on the left side of the bar charts. The
tab provides a short description of the discovered bottleneck and of the state of the pipeline used. The
3D Pipeline
tab of the
pane provides detailed descriptions of discovered bottlenecks and potential ways to fix them.
3D Pipeline
tab blocks with bottlenecks are underlined by red and yellow.
If the Hotspot mode is enabled, the
Full Frame
parameter in the
Frame Statistics
pane shows the aggregate duration of events, ignoring parallel execution.
In the Hotspot mode, calls can be grouped by:
    • Bottleneck/State (default) - calls are simultaneously grouped by two parameters:
      bottleneck description and pipeline state.
    • Bottleneck - calls are first grouped by bottleneck descriptions, followed by pipeline state.
    • State - calls are first grouped by pipeline state, followed by bottleneck description.
  • X
    drop-down button - enables you to choose the available quantitative metrics for the X axis to visualize specific aspects of performance in the frame.
  • Y
    drop-down button - enables you to choose the available metrics for the Y axis to visualize specific aspects of performance in the frame.
  • Group by
    drop-down button - groups the events visualized on the bar charts by:
    • Debug Regions
      - groups of events that were combined and named by a developer.
    • Draw Calls
      - no grouping, each bar corresponds to a single event.
    • Render Targets
      - groups of events that were combined by bound render targets.
    • Shader Sets
      - groups of events that were combined by bound shaders (applies to DirectX* 11, DirectX 12, and Vulkan frames).
    • Command Lists
      - groups of events that were combined by command lists and bundles (applies to DirectX 12 and Vulkan frames).
    • Pipeline States
      - groups of events that were combined by bound pipeline state objects (applies to DirectX 12 and Vulkan frames).
    • Custom Regions
      - enables you to group and name regions in the API Log pane. To learn how to create custom regions, see the API Log pane description.
  • Chart Regions
    drop-down button - enables a markup on the top of the
    bar chart to highlight the same groups that are available in the
    Group By
    drop-down list.
  • Color by
    drop-down button - colors bars on the
    bar chart according to the types of events or bottlenecks. The button is available only if the system supports the full list of Intel performance counters.
  • Import
    button - imports the saved metrics data onto the chart for comparison with the current metrics values. You can only import metrics data for the same frame file with the same set of metrics selected. If you import one or more CSV files into the bar chart, the X-axis is reset to a constant metric.
  • Export
    button - saves metrics data in the CSV format.
You can import and export metrics data in CSV format to compare performance results for the same frame measured with different drivers or hardware settings:
You cannot compare metrics data collected on different platforms with different sets of metrics. Hover over the rectangles above the chart  to view the *.csv filenames and associate the color code in the bar chart with the imported files. To remove the imported metrics data from the chart, click the corresponding rectangle.
  • Show Color Legend
    button - explains the color legend used in the
    bar chart to denote events and bottlenecks. Legend for bottlenecks is available only if the system supports the full list of Intel performance counters. By default, the bars on the
    bar chart are colored by event types.
Main Bar Chart
bar chart enables you to review individual events or groups of events and their contribution to the overall frame time.
An event is any graphics API function that produces a GPU activity, such as "draw" calls or "clear" calls. Once you load the capture file, events are visualized as vertical bars, which can be of one of the following types:
• Draw Call (colored Dark Blue)
• Dispatch Call (colored Purple)
• Copy Call (colored Cyan)
• Clear Call (colored Green)
• Sync Call (colored Yellow) - DirectX12 and Vulkan frames
• RenderPass Call (colored Magenta) - DirectX 12 and Vulkan frames
If you filter the API calls in the API log, the excluded calls are displayed in gray.
Events are shown in the GPU execution order starting from the left. By default, the height of an event bar indicates the duration of the event execution on the GPU. GPU time is measured in microseconds.
You can change the vertical scaling of the bar chart by increasing or decreasing the relative height of the bars using the or  buttons respectively. Alternatively, you can hold
and scroll up or down to re-scale the bars. By default, all bars are displayed in their original height.
To enable or disable the vertical auto-scaling, click the / toggle button. By default, the vertical auto-scaling is enabled.
You can identify all events using a specific texture, buffer, shader, or render target selected in
Resource List
. You can see a colored stripe below each event using the selected resource. The color of the stripe depends on the group the resource belongs to: orange for
, green for
, and purple for
resources. Half-transparent colored stripes indicate that the resource is used by some of the events in the render target.
To analyze metrics data and resources for specific parts of the frame, select a single event or a group of events in the bar chart . The selected bars are highlighted in orange. You can use the Ctrl+Z/Ctrl+Y keyboard shortcuts to return to the previously selected event and back.
If you select one or more events in the chart and then group them by any criterion, for instance, by render targets, this partial selection within the render target is marked with orange hatching:
Frame Navigation Bar Chart
Use the
Frame Navigation
bar chart to navigate along the frame and to zoom into events that will be shown on the
bar chart:
  • To change the horizontal scaling of the
    bar chart, stretch or shrink the
    Frame Navigation
    bar chart scrollbar by dragging its boarders. The
    Frame Navigation
    bar chart provides an overview of the entire frame, while the scrollbar reflects the part of the frame currently displayed in the
    bar chart. If the X-axis represents a non-constant metric, you can double-click the slider to toggle between the full frame view and the currently selected part.

Product and Performance Information


Performance varies by use, configuration and other factors. Learn more at