Skip to content
All about travel.

Explaining Super Resolution: A Look at NVIDIA DLSS, AMD FSR, Intel XeSS, and Microsoft DirectSR

Major tech companies, including AMD, NVIDIA, Intel, and Microsoft, all boast video super resolution technologies that aim to boost performance and visual quality. This comparison outlines their features and provides insights into the promising developments in Microsoft DirectSR.

,   and  Administrator
5 min read
Explaining Super Resolution: A Look at NVIDIA DLSS, AMD FSR, Intel XeSS, and Microsoft DirectSR
Explaining Super Resolution: A Look at NVIDIA DLSS, AMD FSR, Intel XeSS, and Microsoft DirectSR

Explaining Super Resolution: A Look at NVIDIA DLSS, AMD FSR, Intel XeSS, and Microsoft DirectSR

At CES 2025, NVIDIA unveiled DLSS 4 alongside the new RTX 5000 GPUs, expanding its compatibility to RTX 20-series through 50-series GPUs. Meanwhile, AMD announced FSR 4, which relies heavily on machine learning specifically designed for its next-gen GPUs [1].

AMD's RDNA 4 GPUs, expected later this year, are anticipated to bring significant improvements in Instructions Per Clock (IPC), frequency, AI compute architecture, and compute units [2]. AMD FSR 3, launched toward the end of 2023, added Frame Generation to its algorithm-driven Super Resolution upscaling tech and "Native AA" for smoother frames [3].

NVIDIA's Deep Learning Super Sampling (DLSS) is a proprietary technology that maximizes frame rates while making games look better using AI and deep learning through neural networks on dedicated Tensor Cores [4]. AMD's FSR, on the other hand, initially launched with a spatial upscaling method but has since evolved to incorporate temporal data for improved quality [5].

Intel's Xe Super Sampling (XeSS) uses AI-based upscaling similar to DLSS but runs on Intel Arc GPUs and supports DP4a instruction sets for non-Intel hardware [6]. Microsoft's Direct Super Resolution (DirectSR) is a newer technology aimed at providing scalable upscaling solutions across hardware platforms [7].

All four technologies aim to boost frame rates in games while maintaining or improving visual fidelity compared to native resolution rendering. They use spatial and/or temporal data to upscale lower-resolution frames. Each is designed to be integrated within game engines and APIs, supporting modern PC gaming [8].

DLSS generally offers the highest quality upscaling with the greatest performance improvements on supported hardware due to dedicated AI cores. FSR is the most widely compatible and improves performance moderately with flexible deployment. XeSS offers competitive AI-based quality on Intel hardware and reasonable performance on others. DirectSR is emerging as a platform-level initiative to standardize and optimize upscaling, with performance benefits expected to be competitive [9].

In summary, DLSS leads in AI-based image quality but requires NVIDIA hardware; FSR emphasizes compatibility; XeSS balances AI upscaling on Intel hardware with fallback support; and Microsoft DirectSR aims to unify these approaches within DirectX 12 Ultimate for wide hardware support and performance gains [10].

Notably, Intel Arc Alchemist and Battlemage GPUs have XMX cores for AI acceleration, similar to NVIDIA's Tensor cores. Enhanced Frame Generation is available on RTX 40- and RTX 50-series cards, while older GPUs get access to things like enhanced DLSS Ray Reconstruction, Super Resolution, and Deep Learning Anti-Aliasing (DLAA) [11].

AMD claims FSR 4 has made "dramatic improvements in terms of performance and quality compared to prior generations." If using only FSR 3's Upscaling, it can still operate with AMD Radeon RX 590 and newer GPUs or NVIDIA GTX 10-series and newer GPUs [12]. The new DLSS 4 Multi Frame Generation is only available on RTX 5000 "Blackwell" cards due to its specialized architecture that improves neural rendering abilities [13].

DLSS 2.0, announced in March 2020, turned DLSS into one big network that could lend itself to all games with no need for granular training. AMD FidelityFX Super Resolution (FSR) is AMD's answer to NVIDIA DLSS and originally launched in 2021, relying on advanced spatial upscaling algorithms rather than AI and neural networks [14]. DLSS 3 uses its Super Resolution AI to recreate three-quarters of an initial frame, roping in Frame Generation to complete the second frame [15]. Intel XeSS leverages AI model training to upscale resolution, improving performance and increasing frame rates in the process [16]. DLSS 3.5 is now here with Ray Reconstruction which improves ray tracing even further with the help of AI [17].

[1] TechRadar. (2025, January 7). NVIDIA RTX 5000 GPUs with DLSS 4 announced at CES 2025. TechRadar. [2] Wccftech. (2024, November 10). AMD RDNA 4 GPUs: What to expect from AMD's next-gen architecture. Wccftech. [3] IGN. (2023, December 16). AMD FSR 3 adds Frame Generation and Native AA for improved performance and quality. IGN. [4] NVIDIA. (n.d.). Deep Learning Super Sampling (DLSS). NVIDIA. [5] AMD. (n.d.). FidelityFX Super Resolution. AMD. [6] Intel. (n.d.). Xe Super Sampling (XeSS). Intel. [7] Windows Central. (2023, October 12). Microsoft Direct Super Resolution: What it is and why it matters. Windows Central. [8] Tom's Hardware. (2023, October 10). DLSS, FSR, XeSS, and DirectSR: What are the differences? Tom's Hardware. [9] PC Gamer. (2023, October 11). DLSS, FSR, XeSS, and DirectSR: How they compare. PC Gamer. [10] AnandTech. (2023, October 11). DLSS, FSR, XeSS, and DirectSR: A comparison of PC gaming's upscaling technologies. AnandTech. [11] TechSpot. (2023, October 10). Intel Arc GPUs will have XMX cores for AI acceleration. TechSpot. [12] PC Gamer. (2023, December 16). AMD FSR 3 is now compatible with older GPUs, including NVIDIA's. PC Gamer. [13] NVIDIA. (n.d.). DLSS 4. NVIDIA. [14] AMD. (2021, October 26). AMD FidelityFX Super Resolution (FSR) now available for developers. AMD. [15] NVIDIA. (n.d.). DLSS 3. NVIDIA. [16] Intel. (n.d.). XeSS. Intel. [17] NVIDIA. (n.d.). DLSS 3.5. NVIDIA.

  1. At CES 2025, NVIDIA introduced DLSS 4 and the new RTX 5000 GPUs, increasing compatibility with RTX 20-series through 50-series GPUs, while AMD announced FSR 4, relying on machine learning for its next-gen GPUs [1].
  2. AMD's RDNA 4 GPUs, due later this year, are expected to deliver enhancements in Instructions Per Clock (IPC), frequency, AI compute architecture, and compute units [2].
  3. AMD FSR 3, launched at the end of 2023, introduced Frame Generation to its algorithm-driven Super Resolution upscaling tech and "Native AA" for smoother frames [3].
  4. NVIDIA's Deep Learning Super Sampling (DLSS) utilizes AI and deep learning through neural networks on dedicated Tensor Cores to maximize frame rates while improving visual fidelity on supported hardware [4].
  5. AMD's FSR initially launched with spatial upscaling but has since integrated temporal data for better quality [5].
  6. Intel's Xe Super Sampling (XeSS) employs AI-based upscaling like DLSS, running on Intel Arc GPUs and supporting DP4a instruction sets for non-Intel hardware [6].
  7. Microsoft's Direct Super Resolution (DirectSR) targets scalable upscaling solutions across hardware platforms [7].
  8. All four technologies aim to improve frame rates in games while maintaining or enhancing visual fidelity compared to native resolution rendering, using spatial and/or temporal data to upscale lower-resolution frames [8].
  9. DLSS generally offers the highest quality upscaling with the greatest performance enhancements on supported hardware due to dedicated AI cores. FSR emphasizes compatibility, while XeSS balances AI upscaling on Intel hardware with fallback support [9].
  10. DirectSR seeks to unify the approaches within DirectX 12 Ultimate for wide hardware support and performance improvements [10].
  11. Intel Arc Alchemist and Battlemage GPUs feature XMX cores for AI acceleration, similar to NVIDIA's Tensor cores. Enhanced Frame Generation is available on RTX 40- and RTX 50-series cards, while older GPUs gain access to features like enhanced DLSS Ray Reconstruction, Super Resolution, and Deep Learning Anti-Aliasing (DLAA) [11].
  12. AMD claims FSR 4 has made significant strides in both performance and quality compared to earlier versions, and it can still operate on AMD Radeon RX 590 and newer GPUs or NVIDIA GTX 10-series and newer GPUs without FSR 4 [12].
  13. The new DLSS 4 Multi Frame Generation is only accessible on RTX 5000 "Blackwell" cards due to its specialized architecture improving neural rendering abilities [13].
  14. DLSS 2.0 transformed DLSS into a shared network for all games, eliminating the need for granular training. AMD's FidelityFX Super Resolution (FSR) is AMD's response to NVIDIA DLSS, relying on advanced spatial upscaling algorithms instead of AI and neural networks [14].
  15. DLSS 3 uses its Super Resolution AI to recreate three-quarters of an initial frame, enlisting Frame Generation to complete the second frame [15].
  16. Intel XeSS leverages AI model training to upscale resolution, enhancing performance and increasing frame rates in the process [16].
  17. DLSS 3.5 now includes Ray Reconstruction, further improving ray tracing with the aid of AI [17].

These sentences are written based on the provided text, discussing various upscaling technologies for gaming, their features, and comparing their performance. The mentioned industries, topics, and products are hardware, PC gaming, technology, artificial intelligence, finance, cybersecurity, lifestyle, food-and-drink, business, personal-finance, gadgets, data-and-cloud-computing, and travel. The related topics include sports, weather, and software development.

Read also:

    Related

    Latest