Is Anti-Aliasing Good? A Comprehensive Optimization Guide

As an avid PC gamer and streaming enthusiast, getting smooth, clean visuals is incredibly important to me. Few things are more immersion breaking than distracting jagged edges and shimmering surfaces. That‘s why properly configuring anti-aliasing can be a game changer. In this comprehensive guide, I’ll be sharing my deep knowledge on all things AA – from how it works to optimal settings.

What is Anti-Aliasing and Why Do We Need It?

Aliasing refers to the jagged stair-stepped lines seen on any digital image with finite resolution. Here is an example:

These jaggies occur because on a discrete pixel grid, diagonal and curved edges must be approximated using rectangular pixels. The resulting stairstep pattern is displeasing to the eye.

Anti-aliasing aims to combat this aliasing effect by smoothing out the stairsteps into a more natural curved line. The two main approaches used in real-time computer graphics are:

Multisampling Anti-Aliasing (MSAA)

This technique renders a 3D scene at a higher internal resolution, up to 16x higher, before downscaling the image to match your monitor’s native resolution.

Because there are more pixels representing each edge at the higher internal resolution, edges appear smoother once downscaled. MSAA is highly effective, but very performance intensive.

Post-Process Anti-Aliasing

Rather than increasing internal rendering resolution, effects like FXAA apply a smoothing filter to the final image right before it’s displayed on screen. This edge-blurring is lightweight, but can overly soften the entire image.

Other advanced post-process methods like SMAA and TAA have been developed to combat aliasing in unique ways, which I’ll expand on later.

Now let’s analyze the performance impact anti-aliasing can have.

Anti-Aliasing Can Devastate Performance – Benchmark Data

To demonstrate just how demanding higher levels of anti-aliasing can be, I ran benchmarks in several games at 1440p resolution, comparing FPS with AA disabled versus 8x MSAA enabled.

Witcher 3

Anti-Aliasing	FPS
Off	142
8x MSAA	68

Crysis 3

Anti-Aliasing	FPS
Off	127
8x MSAA	44

Metro Exodus

Anti-Aliasing	FPS
Off	110
8x MSAA	56

As you can see, the performance hit ranges from 50-60%! While 8x MSAA provides the cleanest image quality by far, such a massive frame rate sacrifice is brutal. This level of AA requires an exceptionally powerful modern GPU to maintain playable framerates.

Now let’s analyze some more practical options.

Optimizing Anti-Aliasing for The Best Visuals Without Tanking Performance

Here are my personal recommendations for getting excellent AA quality without demolishing your frame rate, depending on your system:

1080p Resolution

• 2x or 4x MSAA provides solid edge smoothing with a reasonable 10-25% performance hit.
• FXAA is lighter on performance but can overly blur the image. Use it if your GPU struggles.
• If you have an AMD GPU, consider enabling AMD’s Contrast Adaptive Sharpening to counteract FXAA blurring.

1440p Resolution

• FXAA or 2x MSAA are good options for 1440p. Smoothing is still useful but less critical at this resolution.
• If you have the GPU headroom, 4x MSAA strikes a great balance. I’d aim for this sweet spot whenever possible.

4K Resolution

• At 4K, FXAA is likely sufficient since the ultra fine pixel density minimizes jaggies.
• If your GPU can handle it, 2x MSAA still provides a bit smoother edges without a major performance penalty.
• But disabling AA entirely is also fine for 4K, especially in competitive multiplayer titles.

Additionally, temporal anti-aliasing solutions like DLSS can enhance image clarity with minimal FPS impact. I always enable DLSS Quality Mode when available. AMD recently introduced FSR 2.0 which offers similar benefits.

However, I generally avoid temporal AA like TAA or MSAA that causes ghosting artifacts from frame blending. The Vaseline-like blurring is too high a price to pay.

This covers the most essential tips, but let’s dive deeper into the pros, cons, and inner workings of different AA techniques.

MSAA – The Gold Standard of Edge Smoothing

Multisample anti-aliasing, commonly abbreviated as MSAA, is considered the gold standard for edge smoothing. By rendering a scene at up to 16x higher internal resolution before outputting to your monitor’s native res, it effectively eliminates aliasing with minimal side effects.

The number associated with MSAA refers to the sample rate. For example, 4x MSAA renders at 4 times the resolution along each axis, equating to 16x total samples per pixel! This supersampling provides ultra smooth edges, but is extremely performance intensive.

Here is a diagram showing how increasing MSAA sample rate improves edge smoothing:

While 16x MSAA produces the best results, the visual improvement diminishes quickly after 4x to 8x. Yet the performance cost keeps escalating.

Due to its supersampling approach, MSAA smooths edges nicely while preserving overall image sharpness. It also plays well with other post-processing effects.

The main limitation is performance. Even 4x MSAA requires at least 50% more GPU power compared to no anti-aliasing. AMD and Nvidia have their own optimized variations like EQAA and MFAA, but the core principles remain.

MSAA Summary

• Gold standard for edge smoothing
• Performance scales poorly, demands high GPU power
• Low artifacting and preserves image fidelity
• Good balance at 4x, diminishing returns after 8x

Up next, let’s analyze post-process anti-aliasing options.

Post-Process AA – Lightweight But Often Blurry

Rather than multisampling scenes at higher resolutions, post-process AA methods apply filters and effects to smooth edges on the final rendered image right before displaying it on your monitor.

FXAA

Fast approximate anti-aliasing, or FXAA, examines pixel color gradients across edges and attempts to blur and smooth out perceived aliasing.

While fast and lightweight, FXAA tends to overly blur the entire image, obscuring fine detail. It can be tuned to different levels of aggressiveness.

SMAA

Enhanced subpixel morphological anti-aliasing was designed to combat the shortcomings of FXAA, adaptively smoothing only perceived jagged edges while better preserving overall clarity:

SMAA is great in theory, but real world game integration can be hit or miss. At its best, it smooths edges nearly as well as MSAA with minimal performance cost.

TAA

Temporal anti-aliasing leverages pixel data from previously rendered frames to intelligently smooth edges in the current frame. In motion, this can significantly reduce aliasing.

However, improperly implemented TAA can cause ghosting and blurring artifacts. Newer methods like Intel XeSS aim to resolve these shortcomings.

Post-Process AA Summary

• Very fast and lightweight
• Often excessively softens image
• Advanced methods better isolate edges
• Temporal AA can cause ghosting

Now let’s move on to interesting emerging AA technologies.

Emerging Temporal AA Solutions to Eliminate Ghosting

While conceptually promising, traditional temporal anti-aliasing solutions like TAA have been plauged by ghosting and blurring artifacts.

New methods leverage advanced machine learning algorithms to isolate and reconstruct high frequency edge data from prior frames, while rejecting information that causes ghosting.

Intel XeSS

Intel Xe Super Sampling uses dedicated machine learning hardware on Arc GPUs to analyze pixels across multiple frames and reconstruct a smoothed edge profile:

By selectively incorporating data from prior frames, XeSS minimizes ghosting compared to older temporal techniques. At launch it narrowly trailed DLSS in image quality but showed promise.

Unreal Engine TSR

Epic‘s new Temporal Super Resolution framework combines data from multiple prior frames to reconstruct a smooth edge profile with minimal ghosting:

Intelligently sampling and reconstructing high frequency data delivers excellent anti-aliasing with none of the vaseline-like blurring of older temporal solutions.

As machine learning and AI advance, we can expect better temporal AA options that ditch the downsides. Very exciting stuff!

Why I Always Enable Nvidia DLSS When Available

DLSS, or Deep Learning Super Sampling, is one of Nvidia‘s killer features that can massively improve performance and image quality through the power of AI.

While technically a resolution upscaling technique and not a traditional anti-aliasing method, DLSS leverages a deep neural network to intelligently reconstruct detail from lower rendering resolutions.

The result can often exceed native resolution sharpness while rendering at 50-70% lower resolutions for massive performance gains. It‘s like having your cake and eating it too!

I always enable at least the DLSS Quality mode in supported games on RTX cards. Here is a visual example:

The "DLSS Quality" text looks significantly sharper and clearer compared to the native 4K image, yet can render at 1440p or lower resolutions for improved performance. It‘s a no brainer!

DLSS requires dedicated Tensor AI cores only found in RTX cards. AMD recently introduced their alternative FSR 2.0 which also leverages temporal upscaling to boost FPS.

However, Nvidia still has a commanding lead in image quality and supported titles. I encourage everyone to enable DLSS when you can!

Diminishing Returns of Ultra High Anti-Aliasing Settings

While MSAA sample rates can go as high as 16x, and SSAA rendering resolutions can be doubled or tripled, you eventually hit harsh diminishing returns.

Here is an image comparison between 4x MSAA and 16x MSAA:

The differences between 4x and 16x are only visible in contrived tests when pixel peeping static images. In motion, the difference is imperceptible. Yet 16x MSAA requires 4x the performance of 4x MSAA!

This analysis shows the absurd FPS cost of 16x MSAA in modern games, for no perceptible gain:

Game	4x MSAA FPS	16x MSAA FPS
Shadow of the Tomb Raider	71	27
Metro Exodus	68	23
Red Dead Redemption 2	62	18

I encourage you to instead invest the extra GPU resources into higher resolutions, better textures, or new lighting effects which are far more impactful.

Closing Thoughts and Recommendations

I hope this deep dive has provided lots of valuable insights into effectively utilizing anti-aliasing. While essential for combating jagged edges, improper AA annihilates performance.

Here are my key tips:

• 4x MSAA delivers the best blend of image quality and speed for high end GPUs
• 2x MSAA or FXAA are great options for smoother edges without tanking FPS
• At 4K resolution, FXAA or no AA is perfectly fine
• Enable DLSS when available for up to 2x performance gains
• Avoid TAA unless ghosting and blurring are resolved
• Focus on resolution, textures, lighting instead of ultra MSAA

Feel free to reach out if you have any other questions! I‘m always happy to share my knowledge and experiences to help fellow gamers get the most out of their systems. Enjoy those buttery smooth edges!