Microsoft has made its own AI upscaler: Auto Super Resolution! Using the NPU in Copilot+ PCs, Auto SR should work on any game - and we've tested it. Here's a look at the feature running on a Surface Laptop 7:

New Xbox console will have this

@XboxP3 Been telling people the next xbox is npu,cpu,gpu! This upscaling is just the tip of what they are working on!

For those wondering what makes this so great compared to how console do some of this today creates such a power suck. The machine learning (ML) capabilities in the Xbox Series X are primarily powered by its custom AMD hardware, which includes a combination of a CPU and GPU designed to support ML tasks. The Xbox Series X uses its GPU, part of the AMD RDNA 2 architecture, for tasks like AI upscaling, enhancing graphics fidelity through techniques like super resolution, and improving overall gaming experiences. Here’s how the ML capabilities of the Xbox Series X compare and differ from an NPU: ### **Xbox Series X ML Capabilities:** 1. **GPU-Centric ML Tasks**: The Xbox Series X leverages its powerful GPU for ML tasks, such as upscaling images to 4K or improving frame rates. This is done using DirectML, a part of DirectX, which allows the GPU to perform machine learning operations. 2. **Versatility**: The GPU in the Xbox Series X is versatile and can handle both graphics rendering and ML tasks. While it’s not as specialized as an NPU, it is capable of a broader range of tasks, which includes real-time ray tracing, advanced shading techniques, and some ML-based enhancements. 3. **Power Consumption**: The GPU is more power-hungry compared to an NPU when performing ML tasks. While it’s optimized for high-performance gaming, it is not as power-efficient as an NPU, which is designed specifically for low-power AI computations. ### **NPU Capabilities:** 1. **Specialization**: NPUs are specialized processors dedicated to handling AI and ML tasks. They excel at performing specific operations like neural network inference with much higher efficiency and lower power consumption than a GPU or CPU. 2. **Efficiency**: NPUs are designed to perform ML tasks with minimal energy usage, making them ideal for mobile and handheld devices where power consumption is a critical concern. 3. **Task Offloading**: NPUs are often used to offload AI tasks from the CPU and GPU, which can lead to significant power savings and performance improvements, especially in devices where battery life is important. ### **Key Differences:** - **Purpose**: The Xbox Series X uses its GPU for a wide range of tasks, including ML, but it’s not as specialized or efficient in ML tasks as an NPU. An NPU, on the other hand, is purpose-built for efficient AI processing. - **Power Efficiency**: NPUs are much more power-efficient than the GPU used in the Xbox Series X for ML tasks. While the Xbox is designed for maximum performance in a home console setting, an NPU is optimized for power efficiency in devices where battery life is crucial. - **Hardware Focus**: The Xbox Series X focuses on delivering high-end gaming performance with its hardware, whereas NPUs are part of a broader strategy to integrate efficient AI capabilities across a range of devices, including handhelds and mobile devices. In summary, while the Xbox Series X is a powerhouse for gaming and can handle ML tasks, an NPU would be far more efficient for AI-specific tasks, especially in devices where power efficiency is paramount.

I guess this is for older games?

I know this is designed with NPUs in mind. BUT, given the GPU grunt of the Series X, couldn't this be implemented into backwards compatible Xbox 360 games? optionally of course given the input lag it adds but that would be a pretty great use case I'd think

@XboxP3 Looks good,

@XboxP3 Awesome technology. Bring it to Xbox baby

I guess I need to watch this in 4k and not on Twitter on my phone cause...

Ok but Sony/Playstation have also created their own AI Upscaler! They did this way before Xbots! Give PlayStation credit now

They look the same unless you pause. How is the performance?