FSR vs DLSS vs XeSS: Complete Upscaling Guide (2026)

DLSS is the best upscaling technology for image quality if you own an Nvidia GPU – but AMD’s FSR works on every graphics card, making it the most accessible option by a wide margin. The FSR vs DLSS vs XeSS debate comes down to three things: what GPU you have, how much image quality you are willing to trade for performance, and whether frame generation matters to you. This guide breaks down exactly how each technology works, where each one falls short, and which one you should enable in your games. If you want the full context, see our complete gaming monitor guide.

Upscaling renders your game at a lower internal resolution, then reconstructs the image to your monitor’s native resolution. The result: higher frame rates without dropping your settings to low. The catch is that no upscaler is free – every one of them introduces artifacts, and the quality gap between DLSS, FSR, and XeSS is significant depending on the mode and game.

What Upscaling Actually Does

Every upscaler follows the same basic idea: render fewer pixels, then reconstruct the missing detail. If you play at 4K (3840×2160), that is 8.3 million pixels per frame. At DLSS Quality mode, the game renders at 1920×1080 internally – just 2 million pixels – then the upscaler fills in the rest. Your GPU does 75% less rendering work, but the final image looks close to native 4K.

“Close” is the key word. No upscaler perfectly reconstructs native resolution. They all guess, and those guesses show up as shimmer on thin geometry, ghosting on fast-moving objects, or soft textures that look slightly out of focus. How well each technology hides these artifacts is where the comparison gets interesting.

There are two broad approaches. Temporal upscaling uses motion data from previous frames to predict detail – DLSS, FSR 2+, and XeSS all do this. AI-based upscaling adds a trained neural network on top of the temporal data – only DLSS uses dedicated Tensor Cores for this, which is why it produces the cleanest results.

DLSS: Nvidia’s AI-Powered Upscaler

DLSS (Deep Learning Super Sampling) runs on dedicated AI hardware inside Nvidia RTX GPUs called Tensor Cores. Nvidia trains a neural network on high-resolution reference images for each supported game, then your GPU runs that trained model in real time. The result is the sharpest, most artifact-free upscaling available.

DLSS has gone through several major versions. DLSS 1.0 was rough – blurry, limited to specific games, and widely criticized. DLSS 2.0 switched to a generalized temporal model and the quality jump was massive. DLSS 3.0 added Frame Generation, which creates entirely new frames between rendered frames. DLSS 4 (early 2026) introduced Multi Frame Generation, inserting up to three AI-generated frames for every rendered frame on RTX 50-series GPUs.

Where DLSS shines: It handles thin geometry (chain-link fences, tree branches, hair strands) better than any competitor. Moving elements stay sharp. Text in HUDs and menus remains readable. At Quality mode, many users cannot distinguish DLSS from native rendering in blind tests.

Where DLSS falls short: It only works on Nvidia RTX GPUs – nothing from AMD or Intel can run it. Game support requires Nvidia to work with developers, so not every title gets it. DLSS Frame Generation adds visible latency unless you also enable Nvidia Reflex, and some users report occasional flickering in particle effects.

FSR: AMD’s Open-Source Alternative

FSR (FidelityFX Super Resolution) is AMD’s answer to DLSS, and it takes a different approach. FSR is open-source and runs on standard shader hardware – no specialized AI cores needed. That means it works on AMD GPUs, Nvidia GPUs, Intel GPUs, and even the PS5 and Xbox Series X.

FSR 1.0 was a simple spatial upscaler – it sharpened a low-resolution image without temporal data. The results were poor. FSR 2.0 added temporal accumulation (using data from previous frames) and closed a large part of the gap with DLSS. FSR 3.0 introduced AMD Fluid Motion Frames, which is AMD’s frame generation competitor. FSR 3.1 improved ghosting reduction and added an API that lets developers swap FSR into games that already support DLSS.

Where FSR shines: Availability. It works on nearly every GPU made in the last five years. Developers can add it with minimal effort because it is open-source. The performance uplift is real – a 40-60% FPS boost at Quality mode is typical. FSR 3 frame generation works on both AMD and Nvidia hardware, unlike DLSS 3 which locks frame generation to RTX 40-series and newer.

Where FSR falls short: Image quality trails DLSS at every preset level. FSR produces more shimmer on thin objects, more ghosting on fast camera movements, and a softer overall image. The gap is smallest at Quality mode and largest at Performance and Ultra Performance modes. Without dedicated AI hardware, FSR is doing more guesswork with less data. AMD’s frame generation also adds more latency than Nvidia’s implementation.

XeSS: Intel’s Middle Ground

XeSS (Xe Super Sampling) launched alongside Intel’s Arc GPUs in late 2022. It uses a hybrid approach: on Intel Arc GPUs, it runs on dedicated XMX (Xe Matrix Extensions) AI cores, similar to how DLSS uses Tensor Cores. On AMD and Nvidia GPUs, it falls back to DP4a shader instructions – better than nothing, but not as good as the hardware-accelerated path.

XeSS 2 arrived in early 2026 with improved temporal stability and Intel’s own frame generation technology. The quality on Intel Arc hardware is competitive with DLSS in many titles. The DP4a fallback path on non-Intel GPUs is noticeably weaker, roughly on par with FSR 2 in image quality.

Where XeSS shines: If you own an Intel Arc GPU, XeSS produces image quality that rivals DLSS. It is also open-source and available cross-platform, giving it wider game support than a locked-down system would allow. XeSS handles motion well and produces minimal ghosting on Intel hardware.

Where XeSS falls short: On non-Intel GPUs, the DP4a fallback produces softer results than DLSS and only marginally better than FSR in some games. Intel Arc GPU market share is small, so game developers prioritize DLSS and FSR support first. XeSS game support is growing but still trails both competitors.

Image Quality Comparison

Quality rankings shift depending on the upscaling preset. At Quality mode (67% render resolution), the gap between DLSS and FSR is small enough that some people do not notice it in motion. Drop to Performance mode (50% render resolution) or Ultra Performance (33%), and DLSS pulls ahead dramatically because the AI model reconstructs detail that the temporal algorithm in FSR cannot recover.

Here is how each technology stacks up across the four standard quality presets:

The biggest visual differences show up on foliage, hair, and distant geometry. DLSS preserves individual leaf edges and hair strands even at Performance mode. FSR at the same preset turns these into a shimmery mess, especially when the camera moves. XeSS on Intel Arc hardware splits the difference – sharper than FSR, slightly softer than DLSS.

One thing every upscaler shares: they all look better at 4K output than 1080p output. Upscaling to 4K from 1080p internal gives you enough pixel density to hide artifacts. Upscaling to 1080p from 720p internal looks rough regardless of which technology you pick.

Performance Impact by Technology

All three upscalers deliver similar FPS gains at the same quality preset because they all render at the same internal resolution. The performance difference between them is small – typically 1-3% – because the bottleneck is the lower-resolution rendering, not the reconstruction step.

Typical FPS improvement at Quality mode (67% render scale):

• DLSS Quality: 40-60% FPS boost. Tensor Core processing adds minimal overhead.
• FSR Quality: 40-55% FPS boost. Slightly less overhead on AMD GPUs where shader scheduling is optimized for it.
• XeSS Quality: 35-55% FPS boost. On Intel Arc, XMX cores handle it efficiently. DP4a fallback on other GPUs costs 2-5% more overhead than DLSS or FSR.

At Performance mode (50% render scale), all three push 70-100% FPS gains. The practical ceiling depends on whether you hit a CPU bottleneck before the GPU has room to breathe.

Frame Generation: DLSS 3/4 vs FSR 3 vs XeSS

Frame generation is the newest and most controversial feature in the upscaling war. Instead of upscaling existing frames, frame generation creates entirely new frames between rendered frames. Your GPU renders frame 1 and frame 3 – the frame generator synthesizes frame 2 from motion data. The result looks like double (or more) the frame rate on your monitor.

The catch: generated frames do not reduce input lag. Your game still only processes input on rendered frames, so your mouse and keyboard responsiveness stays tied to the base frame rate. A game running at 60 FPS with frame generation showing 120 FPS on the counter still feels like 60 FPS for input response.

DLSS 3 Frame Generation (RTX 40-series): Inserts one generated frame per rendered frame, doubling the displayed frame rate. Requires Nvidia Reflex to offset the latency penalty. Only works on RTX 4060 and newer. The generated frames are high quality – most people cannot spot them in motion.

DLSS 4 Multi Frame Generation (RTX 50-series): Inserts up to three generated frames per rendered frame, effectively quadrupling displayed frame rate. An RTX 5070 rendering at 30 FPS can display 120 FPS on screen. The technology requires RTX 5060 or newer and pushes hard into the territory where input lag becomes a real concern for competitive players.

FSR 3 Fluid Motion Frames: AMD’s frame generation runs on standard shader hardware, so it works on both AMD and Nvidia GPUs. This is a major advantage over DLSS 3/4, which locks frame generation to Nvidia hardware. The quality is slightly lower than DLSS frame generation – occasional warping artifacts appear near screen edges and around fast-moving objects.

XeSS Frame Generation: Intel added frame generation with XeSS 2. It works on Intel Arc hardware and produces clean results there. Cross-platform support is limited compared to FSR 3.

Latency and Input Lag Impact

This is where upscaling technologies get messy for competitive gamers. Standard upscaling (without frame generation) adds minimal latency – roughly 1-2ms on DLSS, 1-3ms on FSR, and 1-3ms on XeSS. That is too small for most humans to perceive.

Frame generation is a different story. Every generated frame introduces one frame’s worth of latency because the display shows a predicted frame while the next real frame renders. At 60 FPS base, that is about 16ms of added latency. At 30 FPS base with Multi Frame Generation, the theoretical latency penalty is even higher.

Nvidia Reflex counters this by reducing the render queue and synchronizing input with frame delivery. With Reflex enabled alongside DLSS 3/4 frame generation, total system latency often drops below what you would see running at native resolution without Reflex. AMD Anti-Lag 2 does something similar but is not as widely supported in games.

Bottom line for competitive play: Use upscaling without frame generation. Set DLSS or FSR to Quality mode and leave frame generation off. You get higher FPS and lower input lag than native rendering. Frame generation is for single-player games where visual smoothness matters more than click-to-pixel response time.

Which GPU Supports Which Technology

The short version: Nvidia owners get the most options. AMD owners are limited to FSR and XeSS (DP4a fallback). Intel Arc owners get the best XeSS experience but miss DLSS entirely. Every GPU made in the last five years can run FSR.

Common Mistakes

Using Ultra Performance Mode for Everyday Gaming

Ultra Performance renders at 33% resolution. At 1440p output, that is an internal resolution of 480p. Even DLSS cannot make 480p look good on a regular basis. Ultra Performance exists for 8K displays and extreme edge cases. Stick to Quality or Balanced mode for daily use.

Enabling Frame Generation in Competitive Multiplayer

Frame generation adds latency. Period. The frame counter shows 200 FPS but your input response matches the base 100 FPS. In Valorant, CS2, or any game where reaction time decides gunfights, frame generation costs you duels. Use standard upscaling and turn frame generation off for ranked play.

Ignoring the Upscaler Sharpening Slider

DLSS, FSR, and XeSS all have a sharpening control. Many games default it to 0 or a low value. Bumping it to 0.3-0.5 removes the soft look that people complain about. Going past 0.7 introduces ringing artifacts around edges. Adjust it per-game – the default is rarely the best setting.

Thinking FSR and DLSS Are the Same Because FPS Gains Match

Both give similar FPS boosts because both render at the same lower resolution. The difference is reconstruction quality. Matching FPS numbers does not mean matching image quality. DLSS at Performance mode looks closer to native than FSR at Quality mode in many titles. Compare screenshots, not frame counters.

Running Upscaling at 1080p Output

Upscaling to 1080p means rendering at 720p (Quality) or 540p (Performance) internally. There are not enough pixels at these resolutions for the reconstruction algorithms to work well. The image turns into a blurry, artifact-filled mess. Upscaling works best at 1440p and higher output resolutions.

Which Upscaler Should You Use?

Here is the straight answer based on your GPU and what you play.

Nvidia RTX 40/50-series owner playing single-player games: Enable DLSS with Frame Generation and Nvidia Reflex. You get the best image quality, the highest frame rate, and Reflex keeps latency in check. This is the gold standard for visual gaming in 2026.

Nvidia RTX 20/30-series owner: Enable DLSS at Quality or Balanced mode. Skip frame generation since your GPU does not support it through DLSS (FSR 3 AFMF is an option but adds noticeable latency on these cards). DLSS 2 at Quality mode still beats FSR and XeSS on image quality.

AMD RX 7000/9000-series owner: Use FSR 3.1 at Quality mode. Turn on Fluid Motion Frames for single-player games if you want the extra visual smoothness, but disable it for competitive play. FSR on AMD hardware runs with slightly less overhead than on Nvidia GPUs.

Intel Arc owner: Use XeSS. On your hardware, the XMX-accelerated path produces results that rival DLSS. FSR works too, but XeSS is specifically optimized for your GPU’s architecture.

Competitive FPS player on any GPU: Use whichever upscaler your GPU supports best (DLSS for Nvidia, FSR for AMD, XeSS for Intel) at Quality mode. Turn off frame generation. Turn on Nvidia Reflex or AMD Anti-Lag 2 if available. Your goal is lowering render time, not inflating the frame counter with generated frames.

Older GPU (GTX 1060, RX 580, etc.): FSR is your only real option. Set it to Quality mode and lower your expectations – these GPUs lack the shader performance for FSR to reconstruct cleanly at aggressive presets. You will get playable frame rates but the image will be noticeably softer than native.

If you only remember one thing: DLSS wins on quality, FSR wins on compatibility. Pick whichever one your GPU supports, use Quality mode, and avoid Ultra Performance unless you are at 4K or 8K output.

2026 update: Gaming hardware evolves quickly. For the latest recommendations across mice, keyboards, monitors, and more, browse our interactive tools and hardware databases which are updated regularly with new releases and price changes.

Frequently Asked Questions

Is DLSS better than FSR for image quality?

DLSS produces better image quality than FSR at every quality preset. The gap is smallest at Quality mode (67% render scale) where both look close to native, and largest at Performance and Ultra Performance modes where DLSS retains detail that FSR cannot reconstruct. The tradeoff is that DLSS only works on Nvidia RTX GPUs, while FSR runs on any graphics card.

Can I use DLSS on an AMD GPU?

No. DLSS requires Nvidia’s Tensor Cores, which only exist on RTX 20, 30, 40, and 50-series GPUs. AMD GPUs cannot run DLSS. Your options on AMD hardware are FSR (best native support) and XeSS (via DP4a shader fallback). FSR 3.1 is the recommended choice for AMD GPU owners.

Does frame generation add input lag?

Yes. Frame generation inserts predicted frames between rendered frames, adding roughly one frame of latency. At 60 FPS base rate, that is about 16ms of extra lag. Nvidia Reflex and AMD Anti-Lag 2 offset this penalty, often bringing total latency below native rendering. For competitive games, disable frame generation entirely and use standard upscaling only.

Which upscaling mode should I use for 1440p gaming?

Quality mode is the best choice for 1440p. It renders at 960p internally and reconstructs to 1440p, giving you a 40-60% FPS boost with minimal visual degradation. Balanced mode works if you need more performance but introduces slight softness. Avoid Performance and Ultra Performance at 1440p – the internal resolution drops too low for clean reconstruction.

Is XeSS worth using on a non-Intel GPU?

XeSS on non-Intel GPUs uses the DP4a shader fallback, which produces results roughly on par with FSR 2 in quality. It is not bad, but it is not the best option for Nvidia or AMD owners. Nvidia users should pick DLSS. AMD users should pick FSR. Only use XeSS on non-Intel hardware if a game supports XeSS but not your preferred upscaler.

Related Guides