DLSS, FSR, and XeSS are GPU upscaling technologies that render a game at a lower internal resolution and then reconstruct it to a higher output resolution, raising frame rate while staying close to native image quality. The difference is hardware. NVIDIA DLSS runs only on RTX cards using their Tensor cores, AMD FSR is open and runs on almost any modern GPU, and Intel XeSS runs best on Intel Arc but works elsewhere through a compatibility mode. As of 2026 the current versions are DLSS 4, FSR 4, and XeSS 2, and all three are now AI based. Your GPU brand usually decides which one you should turn on.
What Are DLSS, FSR, and XeSS?
They are three upscaling technologies that let a graphics card produce a high-resolution image from less rendering work. Each one renders the game internally at a lower resolution, then uses the current frame, previous frames, and motion data to reconstruct a sharper image at the resolution shown on screen. The payoff is a higher frame rate at a given output resolution, which is what makes demanding settings such as ray tracing or 4K gaming playable on mainstream hardware.
- DLSS (Deep Learning Super Sampling): NVIDIA’s upscaler, exclusive to RTX graphics cards.
- FSR (FidelityFX Super Resolution): AMD’s open upscaler, designed to run on graphics cards from any vendor.
- XeSS (Xe Super Sampling): Intel’s upscaler, tuned for Intel Arc but able to run on most modern GPUs.
How Does NVIDIA DLSS Work?
DLSS runs a trained neural network on the dedicated Tensor cores inside NVIDIA RTX graphics cards. It takes the low-resolution frame the GPU just rendered, combines it with motion vectors and frames from earlier in the scene, and the network predicts a clean high-resolution result with anti-aliasing built in. Because the work runs on hardware that other vendors do not have, DLSS is exclusive to NVIDIA RTX cards.
The current generation is DLSS 4, which moved to a transformer model, the same class of AI architecture behind large language models. NVIDIA reports it gives better detail in motion, less ghosting, and a more stable image than the earlier convolutional model. The transformer upgrade for Super Resolution and Ray Reconstruction is available across RTX 20, 30, 40, and 50 series cards, while the new Multi Frame Generation is limited to the RTX 50 series.
How Does AMD FSR Work?
FSR is AMD’s open upscaler, and with FSR 4 it became AI based rather than a hand-tuned algorithm. Earlier versions up to FSR 3.1 used a spatial and temporal technique that ran without any dedicated AI hardware, which is why FSR has always worked across a wide range of cards from AMD, NVIDIA, and Intel. FSR 4 changes the approach: it uses a machine-learning model for reconstruction, which closes most of the quality gap with DLSS.
- FSR 4 (AI based): launched on the Radeon RX 9000 (RDNA 4) family, which has the hardware the model needs.
- Wider rollout: AMD is extending the machine-learning upscaling to older RX 7000 cards through a 2026 update.
- Older cards: Radeon cards without that support continue to use FSR 3.1, which still runs without dedicated AI hardware.
- Open source: FSR remains vendor neutral, so it is the upscaler you reach for on a non-NVIDIA card.
How Does Intel XeSS Work?
XeSS is Intel’s AI upscaler, and it runs in two modes depending on the hardware. On Intel Arc graphics cards it uses the built-in XMX matrix engines for the fastest, highest-quality result. On other GPUs it falls back to a compatibility mode that runs the same idea through standard DP4a integer instructions, which most modern cards from NVIDIA and AMD support. This is how an Intel technology runs well beyond Intel hardware, though it looks its best on Arc.
The current generation, XeSS 2, adds frame generation (XeSS-FG) and a low-latency mode called XeLL. Intel has also opened these features to compatible NVIDIA and AMD GPUs, so XeSS has become a genuine cross-vendor option rather than an Arc-only one.
DLSS vs FSR vs XeSS: A Side-by-Side Look
The simplest way to choose is by GPU brand, because that decides which upscaler is available and which one looks best.
What Are the Image Quality and Performance Differences?
At the Quality preset in 2026, DLSS leads, FSR is a close second, and XeSS trails slightly on non-Intel hardware. All three raise frame rate by similar amounts at the same preset, since they all render at the same lower internal resolution. The visible difference is in reconstruction: how clean fine detail looks and how stable the image stays when the camera moves fast.
- DLSS 4: the quality leader. It keeps fine edges, hair, and foliage cleanest and shows the fewest artifacts around moving objects.
- FSR 4: the most improved. On RDNA 4 it is now close to DLSS and far ahead of the older FSR 3.1 on detail and stability.
- XeSS 2: strong on Intel Arc where it uses XMX engines, but it trails on temporal stability when running through the DP4a compatibility mode.
- Presets: Quality renders at a higher internal resolution and looks best; Performance pushes it lower for more frames at some cost to sharpness.
Which Hardware and Games Support Each Technology?
Hardware support is the dividing line: DLSS is locked to NVIDIA RTX, while FSR and XeSS are open to most modern cards.
- DLSS: NVIDIA RTX cards only (the 20, 30, 40, and 50 series). Multi Frame Generation needs the RTX 50 series.
- FSR: the open lineage runs on AMD, NVIDIA, and Intel cards; the AI based FSR 4 needs Radeon RX 9000, with wider rollout in progress.
- XeSS: best on Intel Arc, with a compatibility mode for most modern NVIDIA and AMD GPUs.
- Game support: most major releases ship with at least one upscaler, and many support all three, so the choice usually comes down to your card.
If you are still deciding on a card, the trade-offs between these features sit alongside the wider buying advice in the PC gaming guide.
What Are the Frame Generation Variants?
Frame generation is a separate feature that creates new frames between rendered ones, and each vendor pairs it with their upscaler. Upscaling reconstructs a sharper image; frame generation inserts extra frames to make motion look smoother. The two are often used together, but they are not the same thing, as explained in what frame generation is.
- NVIDIA DLSS Multi Frame Generation: generates up to three frames per rendered frame for a 4X result, exclusive to the RTX 50 series.
- AMD FSR Frame Generation: part of the FSR 4 suite on RDNA 4, with a frame-generation option that also reaches older cards.
- Intel XeSS Frame Generation: XeSS-FG, paired with the XeLL low-latency mode and opened to compatible NVIDIA and AMD GPUs.
- Latency: generated frames raise the number shown on screen, so all three add a low-latency mode to keep the game feeling responsive.
| Feature | NVIDIA DLSS 4 | AMD FSR 4 | Intel XeSS 2 |
|---|---|---|---|
| Method | AI transformer model on Tensor cores | AI / machine-learning model | AI model (XMX or DP4a) |
| Hardware | NVIDIA RTX only (20/30/40/50 series) | Best on Radeon RX 9000 (RDNA 4); broad lineage support | Best on Intel Arc; runs on most modern GPUs |
| Image quality | Highest (sharpest, most stable in motion) | Close second, the most improved | Strong on Arc, trails the other two elsewhere |
| Frame generation | Multi Frame Generation, up to 4X (RTX 50 only) | FSR Frame Generation (RDNA 4) | XeSS Frame Generation with XeLL low latency |
| Openness | Proprietary to NVIDIA | Open source, vendor neutral | Open SDK, cross vendor |
Last Thoughts on DLSS vs FSR vs XeSS
DLSS, FSR, and XeSS solve the same problem in three ways: render at a lower internal resolution, then reconstruct a high-resolution image to raise frame rate. As of 2026 the current versions are DLSS 4, FSR 4, and XeSS 2, and all three are now AI based. DLSS leads on image quality but runs only on NVIDIA RTX cards, FSR 4 is a close second and the most open, and XeSS 2 is strong on Intel Arc with a compatibility mode elsewhere. In practice your GPU brand makes the decision for you.
Upscaling is one of the biggest levers on real-world performance, which is why it sits at the center of any modern build conversation. The hub on PC gaming connects it to frame rate, resolution, and the rest of what shapes how a game runs.
Key Takeaways:
- DLSS, FSR, and XeSS all render at a lower internal resolution and reconstruct it to a higher output resolution.
- DLSS runs only on NVIDIA RTX cards; FSR is open to almost any GPU; XeSS is best on Intel Arc with a compatibility mode elsewhere.
- As of 2026 the current versions are DLSS 4, FSR 4, and XeSS 2, and all three are AI based.
- On image quality, DLSS 4 leads, FSR 4 is a close second, and XeSS 2 trails slightly on non-Intel hardware.
- Frame generation is a separate feature that inserts new frames; DLSS Multi Frame Generation reaches 4X on the RTX 50 series.
- Your GPU brand usually decides which upscaler you should turn on.
Frequently Asked Questions (FAQs)
What is the difference between DLSS, FSR, and XeSS?
All three render a game at a lower internal resolution and reconstruct it to a higher output resolution to raise frame rate. The difference is hardware: DLSS runs only on NVIDIA RTX cards, FSR is open and runs on almost any modern GPU, and XeSS runs best on Intel Arc but works elsewhere through a compatibility mode.
Which upscaler has the best image quality in 2026?
DLSS 4 with its transformer model leads on image quality, with the sharpest detail and the most stable image in motion. AMD FSR 4 is a close second and the most improved, now that it is AI based on RDNA 4 cards. Intel XeSS 2 is strong on Arc hardware but trails the other two on other GPUs.
Can I use DLSS on an AMD or Intel graphics card?
No. DLSS requires the Tensor cores built into NVIDIA RTX graphics cards, so it does not run on AMD or Intel GPUs. On those cards you use FSR, which is open to all vendors, or XeSS, which has a cross vendor compatibility mode.
Is FSR 4 only for new AMD cards?
The AI based FSR 4 launched on the Radeon RX 9000 (RDNA 4) family. AMD is extending its machine-learning upscaling to older RX 7000 cards through a 4.1 update in 2026, while earlier Radeon cards continue to use the older FSR 3.1, which works without dedicated AI hardware.
Does upscaling reduce image quality?
Modern upscalers aim to match native rendering, and at the Quality preset DLSS 4 and FSR 4 often look very close to native while running much faster. Lower presets such as Performance push the internal resolution down further, which trades some sharpness and stability for higher frame rates.
What is frame generation and is it the same as upscaling?
No. Upscaling reconstructs a higher-resolution image from a lower-resolution render. Frame generation creates entirely new frames between rendered ones to raise the displayed frame rate. All three vendors now pair the two, and DLSS 4 Multi Frame Generation can insert up to three frames per rendered frame on RTX 50 cards.
