Instancing and Performance Best Practices

As you add props and other elements to your factory scene, you’ve likely noticed something: many objects are identical. Safety cones, pallets, mounting brackets, warning signs, these items appear multiple times throughout a real facility.

Repeated references can increase the cost of composing and processing a scene. They do not necessarily duplicate every geometry or material buffer; source layers and renderer resources may already be shared.

What Are Instances in USD?

USD scenegraph instancing shares a composed descendant hierarchy through a prototype among compatible instances. Each instance still has its own root and associated overhead. USD determines prototype sharing from composition; merely referencing the same filename does not guarantee compatibility.

Think of instancing like a blueprint system:

• The prototype is the architectural blueprint—one detailed plan

• Each instance is a building constructed from that blueprint—sharing the same design but existing in different locations

How Instances Work Differently

Ordinary References	USD Instances
Separate composed descendant hierarchies	Compatible instances share a prototype hierarchy
Descendants can have per-reference overrides	Descendants cannot be overridden independently per instance
Resource sharing depends on composition and the renderer	Shared hierarchy reduces composition work; per-instance costs remain

See OpenUSD scenegraph instancing for sharing rules and editing restrictions.

Visual Identification in Omniverse

USD instances display distinct visual cues in the Stage panel:

Instance icon Small overlapping I symbol next to the primitive name

Why Instances Are Critical for Digital Twins

Understanding why instancing matters helps you build scalable, performant digital twins that can grow with your facility’s needs.

Performance at Interactive Speeds

Instancing can reduce scene composition and processing costs. Rendering benefits depend on the renderer, asset complexity, and scene configuration; instancing does not guarantee interactive frame rates.

When to Use Instances

Not every object benefits from instancing. Understanding when to apply this technique ensures optimal results.

Ideal Candidates for Instancing

Asset Type	Examples	Why They Work Well
Safety Equipment	Cones, barriers, signs, emergency equipment	Standardized designs, multiple locations, regulatory compliance
Infrastructure Components	Mounting brackets, cable supports, pipe fittings	Standardized parts, repeated throughout facility, identical specifications
Environmental Elements	Lighting fixtures, flooring tiles, structural beams	Consistent appearance requirements, multiple installations
Transport and Storage	Pallets, containers, conveyor components	Standardized dimensions, frequent repetition, material handling needs

When Not to Use Instancing

Avoid For	Reason	Alternative Approach
Unique machinery	Each machine serves different functions, has different configurations	Use references with individual customization
Highly customized objects	Objects requiring significant individual modifications	Create variants or separate asset versions
Single or few objects	Savings may be too small to justify editing restrictions; measure first	Standard referencing is sufficient
Frequently changing objects	Objects that need regular individual updates	Standard references allow easier individual editing

Decision Framework

Use this simple framework to determine if instancing is appropriate:

1. Count Do you have repeated compatible objects? There is no universal minimum count for a benefit.

2. Complexity Is the object geometrically complex enough to benefit from shared memory?

3. Stability Will the objects remain largely unchanged once placed?

4. Scale Do you expect to add more identical objects in the future?

If these conditions apply, compare the same scene with and without instancing. Record instance/prototype counts, memory use, and viewport frame time under the same loading and viewing conditions before deciding.

Getting Started: Preparing for Instancing

Before creating instances, prepare your scene and identify appropriate candidates from the props you’ve already placed.

Identifying Instance Candidates

Look through your current factory scene for:

• Multiple identical props placed in different locations

• Objects you’ll likely need more of as the scene develops

• Standardized components that appear throughout industrial facilities

Step-by-Step Process: Creating Your First Instance

We’ll walk through converting existing props to instances, then creating additional instances from the prototype.

Converting Existing References to Instances

Starting with props you placed in previous sections:

Creating Instances From the Properties Panel

1. Select an object you want to instance

2. Open the Property panel

3. Toggle “Instanceable”

Monitoring Your Instancing Success

Omniverse provides tools to verify that instancing is working correctly and delivering performance benefits.

Enabling USD Stage Statistics

To monitor instancing effectiveness, enable the USD Stage Stats extension:

4. Navigate to Developer > Extensions

5. Search for “USD Stage Stats”

6. Enable the extension

7. Close the Extensions window

Opening the Statistics Window

1. Open Window > Utilities “Statistics” to open the statistics panel

2. In the Statistics window, set “Scope to Display” to “USD Stage”

Understanding the Statistics

The USD Stage statistics display critical instancing metrics:

Metric	What It Means	Why It Matters
PrototypeCount	Number of unique prototypes in your scene	Shows how many different objects you’re instancing
TotalInstanceCount	Total number of instances across all prototypes	Indicates the scale of your instancing implementation
TotalPrimCount	Overall primitive count in the scene	Helps understand scene complexity and optimization opportunities

Interpreting Your Results

Good instancing implementation shows:

• High instance-to-prototype ratio Many instances per prototype indicates effective reuse

• Controlled primitive growth Total primitive count grows slower than object count

• Memory efficiency Compare memory growth against the same number of ordinary references; each instance still adds overhead

Example interpretation:

This is different than the screenshot above.

PrototypeCount: 5
TotalInstanceCount: 25
TotalPrimCount: 150

This shows 5 prototypes supporting 25 instances—an average of 5 instances per prototype. These counts indicate sharing, but do not by themselves prove a memory or frame-time improvement.

The Bigger Picture: Why This Matters

While our current factory scene may not show dramatic performance improvements yet, instancing becomes absolutely critical as your digital twin scales. Here’s why establishing these practices now sets you up for success:

Small Scenes vs. Large Facilities

Current Scene Impact With 10–20 props, measure whether instancing provides a useful benefit for the assets and renderer in use.

Industrial Scale Impact Real manufacturing facilities contain thousands of repetitive elements.

Future Capabilities Enabled

Proper instancing practices today enable advanced capabilities tomorrow:

• Simulation Integration Shared scene structure can support larger simulations; physics costs still require separate measurement

• IoT Data Integration Sensor data from hundreds of devices can be efficiently processed

• Collaborative Workflows Shared assets can support team workflows; collaboration and performance still depend on the application and scene

• Real-time Updates Measure how quickly facility changes appear with the chosen scene structure

The habits you build with these initial props, identifying repetitive elements, converting them to instances, and monitoring performance, become the foundation for managing digital twins at industrial scale. Every safety cone you instance now is practice for the thousands of components you’ll manage in production environments.

Remember: the goal isn’t just to complete this exercise, but to develop the instancing mindset that makes complex digital twins possible. Your facility’s future digital transformation depends on these fundamental optimization practices.