Ultra Impact Shaders

Ultra impact is the top of the site and the smallest group: five packs, every one of them built around light that is traced rather than approximated. SEUS PTGI HRR, Photon Shader and Rethinking Voxels ask for 6 GB of video memory. Continuum Shaders and Hysteria Shaders ask for 8 GB. All five want 8 GB of system memory behind them.

All five come from the high-end category, which makes this the only tier on the site with a single source. What you are paying for is light that comes from somewhere: a torch lights the wall it hangs on in the colour of its own flame, and sunlight through a window bounces off the floor onto the ceiling. A screen-space pack cannot do the second half of that, because the light has to be on screen already.

Version support is the first thing to check here and the most common reason people give up on the tier. SEUS PTGI HRR runs to 1.20.4 and Continuum Shaders to 1.20.4 as well, so neither reaches 1.21. Photon Shader covers 1.19 to 26.1.2 and Rethinking Voxels 1.19 to 1.21.10, and Hysteria Shaders needs 1.20 or newer and ends at 1.21.3.

SEUS PTGI HRR is the pack that made this tier a category of its own, with path tracing, ray-traced reflections and shadows, and underwater caustics that nothing screen-space can reproduce. It is OptiFine only and stops at 1.20.4.

Photon Shader and Rethinking Voxels are the two realistic options for a 1.21 world, both Iris only. Photon does path-traced global illumination with a neutral grade. Rethinking Voxels builds a voxel model of the scene, so coloured light stays correct around a corner instead of only where you are looking.

Continuum Shaders and Hysteria Shaders are the two that ask for 8 GB of video memory. Continuum is the older of the pair and stops at 1.20.4; Hysteria combines voxel global illumination with path tracing and needs 1.20 or newer. Neither is a pack for a normal survival session on a mid-range card.

The memory numbers on this page are not padding. Tracing light through a scene means keeping a version of that scene available to trace against, which is data the game itself never needed to hold. On top of that comes the accumulated result of previous frames, which is how these packs get a clean image without tracing an impossible number of rays per frame. Both scale with render distance, which is why the same pack can be comfortable at eight chunks and impossible at sixteen on identical hardware.

That accumulation across frames also explains the one artefact everyone notices and nobody can name. Move the camera quickly and the lighting takes a moment to settle, leaving a faint trail behind bright objects. It is not a bug and it is not your card failing. It is the pack reusing what it worked out last frame because working it out again from scratch would cost more than you have.

Practically, the levers here are the ray count and the render distance, in that order. Lowering the number of rays makes the image noisier before it makes it slower, and the accumulation hides most of that noise while you are standing still. Lowering render distance gives back memory, which is what usually gives out first at this level.

It is also worth being honest about what this tier is for. Five packs, all of them slower than anything else on the site, two of them stopping at Minecraft 1.20.4. If the goal is a world you play in every evening, the level below this one gets most of the way there for half the memory. If the goal is a frame you are going to keep, this is where you come.

Most people who load an ultra pack are making something: a build tour, a thumbnail, a cinematic. That use has its own rules, and they are different from the rules for playing.

Frame pacing matters more than frame rate. A recording at a steady thirty frames looks better than one that swings between forty and sixty, because the swing shows up as stutter in the finished video and the average does not. Cap the frame rate below what the pack can hold rather than letting it run free.

Give the lighting time to settle before you start moving. These packs build their image over several frames, so the first moments after you turn or teleport are the noisiest. A shot that begins two seconds after the camera has stopped is cleaner than the same shot started immediately, and it costs nothing but patience.

Move slowly and turn slowly. Fast camera movement is the worst case for every pack on this page, because it throws away the accumulated lighting from previous frames. The same scene filmed with a slow pan looks dramatically better than one filmed with a quick turn, on identical hardware.

Set the time of day deliberately. Traced light is at its most convincing when it has something to bounce off at a low angle, which is why early morning and late afternoon flatter these packs and midday flattens them. If a build looks disappointing under an ultra pack, the hour is as likely to be the cause as the settings.

Finally, keep a lighter pack installed alongside. Building under one of these five and recording under it are different activities, and switching to a medium-impact pack while you place blocks costs a few seconds and saves an evening of stutter.

New packs appear at the lighter end of this site regularly. At this end they appear rarely, and the ones that exist have been around for years. That is not a lack of interest; it follows from what these packs are attempting.

Tracing light through a scene is a research problem as much as an engineering one, and the hard part is not the tracing but the noise. There is never budget for enough rays, so the image has to be reconstructed from too little information, and every pack here is really a different answer to that reconstruction problem. Answers of that kind take years to develop and do not survive being rushed.

It also explains why this end of the field is the first to break on a game update and the slowest to come back. A pack that reads the world in detail depends on internal specifics that a lighter pack never touches, and when one of them moves, the fix is not a patch but a redesign of one part.

The practical consequence for you is patience with a deadline. If a pack here has not been updated in a year, treat it as finished rather than as pending, and pick from what is currently moving. If it has been updated recently, expect it to lag a few weeks behind each new game version, and plan your updates around that rather than against it.

Numbers for packs at this level travel widely and are almost never comparable. Three questions decide whether a figure means anything for you.

In what world, and standing where. A flat plains view and a forest above a cave system are different workloads by a wide margin, and the second is where these packs are least comfortable.

At what render distance and resolution. Both scale the cost heavily, and neither is mentioned in most videos. A figure without them describes one person's afternoon rather than the pack.

And measured over how long. A twenty second capture of a stationary camera tells you nothing about the twenty minutes of play that follow it, which is when thermals, memory pressure and terrain loading actually decide whether the session is enjoyable.

If a pack here stalls on your machine, try a shorter render distance before you try anything else. Almost everything expensive at this level scales with how much of the world has to be kept available, and that single setting decides more than all the quality options put together.

Run the same world with the shader off and fly across unexplored ground. If that is uneven, the limit you are hitting is not the pack, and nothing on this page will behave until that is sorted. Chunk generation is work for the processor and the disk rather than the graphics card, and it is the most common thing mistaken for a shader being too heavy.

Every figure on this page is compiled per pack rather than copied from a summary, and where a pack ships several presets the figures describe the one its author treats as the default. That matters most at this level, because the gap between a pack's lightest preset and its heaviest is wider here than anywhere else on the site, and a comparison that quietly mixes the two is worse than no comparison at all.

If something here is wrong, or a pack has changed hands, the support channels linked at the bottom of every page reach us and corrections get acted on.

The first is to decide before you load what the session is for. Building, exploring and filming want different things from a pack at this level, and the mistake nearly everyone makes is to pick the heaviest option once and then use it for all three. Loading takes seconds; regretting it takes an evening.

The second is to leave the frame rate uncapped only when you are testing. For anything else, cap it a little below what your machine comfortably holds. You lose a number you could not feel and you gain a steadier image, a cooler card and a quieter room, which over a long session is the difference between a setup you keep and one you quietly stop loading.

Neither habit is about settings inside the pack, and both do more for how this level feels than any slider in it.

There is a third that only applies here. Keep a note of what you changed and why, because the settings menus at this level are long and the names are not standardised between packs. Six months later, when you come back to a world after an update has reset something, the difference between a working evening and a lost one is usually a sentence you wrote down and forgot about.