The Arm Neural Technology story so far
Arm Neural Technology uses machine learning to reconstruct pixels, generate frames, and make advanced rendering workloads practical within mobile power and thermal limits.
You already know the following technologies:
- Neural Super Sampling (NSS)
- Neural Frame Rate Upscaling (NFRU)
- Neural Super Sampling and Denoising (NSSD)
You can use Arm Neural Technology Playbook - Evaluate to decide whether these techniques fit your game. The evaluate playbook explains what Arm and Sumo Digital learned from Neural Dawn and compares the opportunities and tradeoffs of NFRU and NSSD. Return to that playbook if you need to revisit the technology or the questions to decide fit.
This playbook starts at the next decision: You see a potential fit, so how do you make the first move?
You don’t need to have read the earlier playbook. The only context you need is that neural graphics isn’t one feature with one integration path.
Your route depends on the following factors:
- What problem you want to solve
- What engine you use
- How deeply you want to change your rendering and model-development workflows
Your actionable starting points are the public NFRU and NSS integration routes. Treat NSSD separately as experimental exploration with Arm through the Early Access Program.
Think of the playbook as a series of decisions. Each step narrows the next move:
| Step | Action | Decision |
|---|---|---|
| Choose your play | Start with your goal and engine | Which route fits your game? |
| Learn NFRU | See what happens between two rendered frames | What must your engine own? |
| Test the result | Try NFRU with Project Moku, then with your game | Does NFRU meet the bar? |
| Look beyond | See what NSSD could make possible for lighting | Is deeper exploration worth it? |
| Prepare model work | Know how to change the model to fit your content | Does your model need to change? |
| Use the model toolkit | Connect Arm ML Inference Advisor (MLIA) estimates to Streamline device measurements | What evidence supports your model choice? |
| Look ahead | Get ready to put your work on neural accelerator (NX) devices | What needs to be ready for deployment? |
Start with the public NFRU and NSS routes
NFRU and NSS both have public integration routes. When Arm first shared NFRU, developers were directed to the Early Access Program. NFRU is now available through the public Arm Neural Graphics SDK for Game Engines. It’s also available through the Arm Neural Graphics plugin for Unreal Engine.
Use the plugin for the shortest Unreal Engine route. Use the open, engine-agnostic SDK when you need direct control in another engine. The release also includes NFRU support in Neural Graphics Model Gym and additional debugging controls. For more information, see the SDK v1.1 release notes and Arm Neural Graphics SDK Developer Guide .
The Introducing NFRU in Arm Neural Graphics SDK for Game Engines and UE plugins blog connects the release to the wider story. Specifically, it describes why NFRU complements NSS, how the SDK and Unreal Engine routes differ, and where to start.
Connect model assessment to runtime profiling
The model toolkit supports decisions before and after you have an NX device. ML Inference Advisor (MLIA) checks supported models for Neural Technology compatibility and estimates NX performance on your development host. These estimates help you compare candidates and aren’t measurements of the frame time or image quality of your game.
Arm Streamline 9.9 adds profiling for neural graphics workloads that use the ML extensions for Vulkan. The Mali Timeline view places Data Graph work beside the rest of the system and separates Mali NX execution from shader execution.
The Dispatch Graph Workloads view then shows which graph operations consumed the time, fell back to shaders, or behaved as outliers. Follow Profile Arm Neural Technology workloads with Streamline for device setup, capture, and analysis. Later, use the model toolkit to decide which evidence you need at each stage.
What remains experimental
NSSD represents a more advanced and demanding use of neural graphics. Arm has built and demonstrated the technology, but this playbook doesn’t present it as a product that you can enable. Treat it as experimental Early Access exploration with Arm. Later, you’ll learn what it can unlock and when that exploration makes sense.
What you’ve learned and what’s next
You’ve learned that you can act now through public NFRU and NSS integration routes. NSSD remains an experimental Early Access exploration with Arm.
Next, you’ll choose the route that matches your goal, engine, and appetite for engineering change.