Generated by DeepSeek V3.2| Xe Super Sampling | |
|---|---|
| Name | Xe Super Sampling |
| Developer | Intel |
| Type | Spatial upscaling |
| Released | 2021 |
| Platform | Intel Arc |
| Predecessor | Deep Learning Super Sampling |
Xe Super Sampling. It is a spatial upscaling and anti-aliasing technology developed by Intel for its Arc series of discrete graphics processing units. The technique leverages machine learning to reconstruct a higher-resolution image from a lower-resolution render, aiming to boost frame rates while maintaining visual fidelity. Its introduction marked a significant entry by Intel into the competitive landscape of real-time graphics enhancement.
Xe Super Sampling is integrated into the Intel Xe-HPG microarchitecture and is a core component of the Intel oneAPI rendering toolkit. The technology was announced alongside the Alchemist (microarchitecture) generation of GPUs. It functions as a driver-level feature within the Intel Graphics Command Center, designed to be accessible across a range of supported DirectX 12 and Vulkan applications. The goal is to provide a performance uplift similar to competing solutions from AMD and NVIDIA.
The process begins with the game engine rendering a scene at a sub-native resolution. Xe Super Sampling then employs a spatial upscaler that uses jittered sampling patterns, akin to techniques like Temporal Anti-Aliasing, to gather pixel data. This data is processed through a machine learning model trained on the OpenVINO toolkit. The model, executing on the Xe Matrix Extensions units of the GPU, reconstructs edge detail and texture information to output an image targeting the display's native resolution. A key differentiator is its open-source Intel XeSS SDK, which allows integration at the engine level.
In performance analyses by outlets like Digital Foundry and Tom's Hardware, Xe Super Sampling has demonstrated substantial frames per second gains in titles such as Shadow of the Tomb Raider and Hitman 3. Image quality is generally considered competitive at its "Quality" mode, though some reviews note artifacts like ghosting or softened textures in high-motion sequences compared to native rendering. The technology offers multiple presets, including Performance, Balanced, and Quality, allowing a tunable balance between speed and fidelity. Its effectiveness is closely tied to the specific optimization within each video game.
Xe Super Sampling is most directly compared to AMD FidelityFX Super Resolution and NVIDIA DLSS. Unlike the proprietary AI accelerators used by NVIDIA Tensor Cores, Xe Super Sampling utilizes both dedicated XMX hardware for Intel Arc GPUs and a fallback DP4a instruction path for compatible hardware from AMD and NVIDIA. This contrasts with the vendor-locked approach of DLSS Super Resolution. In terms of adoption, it trails the widespread integration seen by FSR but offers a more open development model than NVIDIA's Deep Learning Super Sampling.
For full hardware acceleration, a GPU based on the Intel Arc Alchemist architecture or later is required. The technology also supports a cross-vendor mode on many modern GPUs from the GeForce 20 series and Radeon RX 6000 series onward via DP4a instructions. On the software side, it requires Windows 10 or Windows 11, a compatible GPU driver, and game integration via the XeSS SDK. The list of supported games includes titles like The Riftbreaker and Call of Duty: Modern Warfare II.
Development of Xe Super Sampling was led by Intel's Accelerated Computing Systems and Graphics Group under Raja Koduri. It was first publicly demonstrated at Intel Architecture Day 2021 as part of the Intel Xe roadmap. The SDK was released to developers in early 2022, with the first game integrations arriving alongside the launch of the Intel Arc A770. Its development signifies Intel's strategic effort to establish a full-stack graphics ecosystem, competing directly with AMD Radeon and NVIDIA GeForce technologies in the PC gaming market.
Category:Computer graphics Category:Intel Category:Graphics processing units