Java Edition distance effects

In, certain game mechanics start to break down as the player reaches the edges of the world.

Entities

 * Projectiles seem to collide incorrectly with entities above 8,388,608 blocks

Rendering

 * Rain and snow appear stretched out at large heights.
 * Translucent blocks can sometimes occlude other translucent blocks behind them depending on player position.
 * This is minor within vanilla's bounds, but becomes very pronounced at much higher distances if using mods, notably over icy areas. Can also affect certain blocks by themselves like slime blocks and honey blocks at even more extreme distances.

Sounds

 * Many break down slightly
 * Becomes considerably more pronounced beyond vanilla bounds at 2 (268 million blocks)
 * The vibration feature for sculk sensors is created at the wrong position

World

 * Temperature distribution breaks at high distances, which can be easily noticed with the creation of snow and ice in biomes such as Mountains appearing blockier due to both world generation and subsequent regeneration from snowfall or freezing
 * As it is set to 60,000,000 by default, which is greater than 33,554,432, it can only be set to a multiple of four in areas near its default position.

Beyond the vanilla world boundary (X/Z ±29,999,984–2,147,483,647)
Horizontal distances far beyond 30 million blocks cannot be reached without game modification. Mods such as the FarLands mod can be used to move the world border further out to make these regions accessible. This lists effects that are completely exclusive to these distances and cannot be seen in any form in vanilla.

Rendering

 * Rain and snow fade at certain horizontal distances.

Entities

 * The player can easily get stuck in the positive sides of blocks after 2 (1.073 billion blocks).

Beyond the 32-bit limit (X/Z ±2,147,483,648-9,223,372,036,854,775,807)
The standard format for doubles dedicates 52 bits to the fraction, as opposed to the 23 bits used by the 32-bit float. As a result, beyond 2^30 or 1,073,741,824 blocks, the player would only be off by (2^30)/(2^52) = 1/2^22 = 1/4194304 blocks, which is absolutely indistinguishable from the distance back at spawn. This is around equivalent to the precision of 2 to 4 blocks out on Bedrock Edition.

Each doubling, however, will indeed half the precision used, up to a point where every single element of the game ends up breaking down.

64-bit precision loss
Minecraft: Java Edition uses 64-bit floating point precision for entity positions and other calculations. Several mechanics which do not break down within vanilla or even slightly modded bounds break down at very high distances similarly to Bedrock Edition.

Entity movement
On the Y axis:
 * Flying upwards or downwards in Creative becomes impossible after 2^52 blocks.
 * Falling downwards becomes impossible after 2^55 blocks.

Stripe Lands
As 52 bits are dedicated to the fraction in the double format rather than 23 in the single format, after 2^53 or 9,007,199,254,740,992 blocks out, precision breaks to consider only every second block, and so on. The rendering breaks down in an effectively identical manner to Bedrock Edition and yields the famous Stripe Lands as a result.

Fluids break down differently from blocks; while block rendering breaks down to form the usual stripes, fluids will instead stretch to the size of the precision loss, with the initiation of the Stripe Lands causing each liquid to become two blocks long, then four at the next doubling, and so on.

Analysis
Due to precision loss becoming more extreme at greater distances, features affected at it will behave different depending on how far out they are.

Rain/snow rendering
First affected bracket: First affected version: Unspecified Classic Last affected version: Indev 2010-02-14 2

Second affected bracket: First affected version: Alpha v1.0.4 Last affected version: Alpha v1.1.2_01

Third affected bracket: First affected version: Beta 1.6.5 Still affects the current release (1.17.1) and snapshot (1.17.1-rc2) Suspected to affect as far back as Beta 1.5, but cannot be reasonably tested due to crashes

Beyond this point on the Y axis one can start to see the first signs of snow/rain jittering. Up to 65,535 blocks. this can only be reasonably seen with snowflakes with a mainly horizontal trajectory, as vertical travelling snowflakes are moving at a speed where travel still appears mostly smooth. Beyond 65,536 and especially 131,072 blocks, the effect becomes very obvious for almost all snow.
 * 16,384 - 262,143 blocks

Above 262,144 blocks, the first signs of geometrical distortion in the snow itself can be seen - very little non-misshapen snow is present beyond here, and most of it has transformed into either lone rectangles, or the odd paired rectangles similar in shape to a pause button.
 * 262,144+ blocks

Deformity progresses after every power of two surpassed from this point. Past 16,777,216 blocks, snow becomes a near unrecognisable pattern of suspended vertical lines.

Translucent rendering breakdown
First affected version: 13w41a Still affects the current release (1.17.1) and snapshot (1.17.1-rc2)

The effects of this can first be seen after this point. The preferred method of testing this involves stacking two translucent blocks, such as two different colors of stained glass, in a square shape, and then walking up against this square to slow movement. When crossing from one block into another, the top face of the bottom block should not render at all until a certain distance is covered, at which point it should pop into view as expected.
 * 8,388,608 - 16,777,215 blocks

The buggy effect's precision is now halved, allowing for the tops of two blocks at a time to appear periodically invisible.
 * 16,777,216 - 33,554,431 blocks

The effect's intensity will again double for every power of two crossed beyond this point. It also becomes very obvious in natural generation at extreme distances, specifically in cold areas; as ice and water are both translucent blocks and are intended to be visible through each other, viewing them when at such far distances from the origin will make very exaggerated versions of this effect obvious.
 * 33,554,432+ blocks

Sound positioning errors
Becomes very severe beyond 2^28 blocks, where many sounds are simply no longer audible at all.

Sculk vibration positioning errors
First affected version: 20w49a Still affects the current release (1.17.1) and snapshot (1.17.1-rc2)

Precision will be a factor of two times as accurate for every power of two crossed before this point. Beyond 4,194,304 blocks, vibrations will only be able to come from one of three lines on a block - one for each edge, and another through the center. Past this on both axes, precision loss will allow vibrations to only originate from nine points on a block: the center, the middle of each of the four edges, and the four corners.
 * 4,194,304 - 8,388,607 blocks

After this point, precision will be half as precise again. Vibrations will lose the ability to come from the center of blocks, being relegated to one of two edges, or, if beyond this on both axes, one of the four corners.
 * 8,388,608 - 16,777,215 blocks

Beyond 16,777,216 blocks, vibrations can come from completely separate positions than the actual source. If a button is placed at (4x+1,4z+1), the vibration will first appear at the center of the four blocks directly northwest of it. However, this does not appear to allow the sculk sensor to detect vibrations from any farther away than it should.
 * 16,777,216 - 33,554,431 blocks

Temperature distribution breakdown
First affected version: 16w02a Still affects the current release (1.17.1) and snapshot (1.17.1-rc2)

As snowfall/rainfall is handled on a per-block basis, the effects of precision loss here can only be seen once precision itself can no longer represent blocks (integers) individually.
 * 16,777,216 - 33,554,431 blocks

Beyond this point, while perhaps not immediately obvious (especially due to the vertical variation in almost all biomes where this effect can be seen), the patterns resulting from snow landing on surfaces become much more angular than before, being commonly composed of large rectangles, thin lines and lone dots which are either filled with snow or have it completely absent. This is similarly true of water, with ice corresponding to cold blocks and water to warmer blocks.

As temperature varies with height, in order to properly see the effects of this, it is strongly recommended to build a flat plane for snow to accumulate on instead, or to generate a Superflat world with snow/ice set to generate with it as it would naturally. A modified Tunneler's Dream preset set to generate 94 layers of black concrete (Looking At Block should say 93 for the top concrete layer) is ideal for this case, providing a roughly 50/50 density of snowy and clear blocks, with black providing maximum contrast.

Teleporting to 16,777,216 on both axes should show four quadrants - one with normal looking snow/ice generation, and three with far more angular features due to the precision loss exceeding a full block. During times of precipitation, it can be seen that the blocky patterns of snow/ice match up with the weather directly above - snowy areas have snowfall where areas with no snow cover have rain. This is obviously true anywhere and is unrelated to precision loss, but (especially in the case of already-generated worlds) this can be used to prove that the precision loss lies with temperature calculation, and is not merely a world generation issue disjoint from it.

History
Note: This history section is sorted by when each case was fixed, shaded areas in gray in the introduction/fix columns are non-actual versions or testing for versions needed, and shaded areas in pink indicate distance effects affecting in the development versions only.

High-time precision loss bugs
While not strictly distance effects, these similarly concern precision loss issues with float casting that happen due to excessive time values, regardless of spatial position.

Redstone wire
First affected version: Alpha v1.0.1 Last affected version: 14w11b Notable changes in Beta 1.5, Beta 1.8 Pre-release and 14w02a

Between Alpha v1.0.1 and Beta 1.7.3 inclusive, the first distortions of redstone wire can be seen beyond 262,144 blocks. After this point, if redstone wire climbs a block face perpendicular to the direction 262,144 blocks are exceeded on, it will appear in the same plane as that block face, resulting in major z-fighting. This effect remains like this until 16,777,216 blocks, where error exceeds a full block.
 * 262,144 - 1,048,575 blocks

In Beta 1.8 Pre-release and later, vertically travelling redstone is no longer affected by precision loss.

Between Alpha v1.0.1 and release 1.8 snapshot 14w11b inclusive, redstone wire first begins to distort beyond 1,048,576 blocks. This distortion is rather subtle, and solely affects the dot - other connections are unaffected. In the screenshot, the redstone at the top (which is the smallest) is before 1,048,576 blocks on both axes, and thus appears at its normal size. The bottom redstone, which is beyond this position, appears bigger. The other two dots are beyond this position on only one axis each, and appear slightly rectangular as a result.
 * 1,048,576 - 4,194,303 blocks

Beyond 4,194,304 blocks, redstone starts to distort more significantly and in more ways. Notably, single dots are invisible - since at these coordinates, precision can only be half a block, this splits a block's top surface into a 2x2 grid with 9 possible vertices. As the redstone dot is centered onto the block, the vertices of the model are presumably rounded to the center of the block, producing an invisible 0-dimensional result. Corner turns of redstone dust are now somewhat shrunken as to only cover a corner of the block face, and T-shapes are stretched to cover only half of the block's top face. Crosses and lines are unaffected.
 * 4,194,304 - 8,388,607 blocks

Redstone rendering degrades further upon another subsequent doubling of the distance. Redstone dots' model vertices are now rounded out to the corners of the block rather than the center, producing an again visible result which now covers the full block face. T-shapes and corner turns now also cover the full block top face. Lines and crosses remain unaffected at this point due to them covering the full top face in the first place.
 * 8,388,608 - 16,777,215 blocks

Tripwire and tripwire hooks
First affected version: 12w22a Last affected version: 1.7.10

Beyond 262,144 blocks on both axes, tripwire string simply stops rendering. Beyond this point on only one axis, it will only stop rendering in the direction perpendicular to the axis it is exceeded on. Of what remains, some subtle z-fighting can be seen.
 * 262,144 - 8,388,607 blocks

Pistons
First affected version: Beta 1.7 Last affected version: 1.7.10 Note: this section deals with stationary pistons and does not include pistons in motion.

The arm section of a piston begins to distort at this point. If the piston's direction is perpendicular to that of the direction of travel (including vertically facing), the arm will now appear completely flat on that axis. If beyond this point on both axes and extending in a vertical direction, this will result in the arm to appear completely 1-dimensional and therefore invisible.
 * 2,097,152 - 4,194,303 blocks

The above details still apply at this point. Pistons now gain an additional distortion at this point. If it is extending parallel to the direction of travel, it will now appear shorter, and detached from the piston body slightly. At 4,194,303 blocks exactly, the texture on the arm will appear slightly squished if facing in the negative direction (as tested on only one positive axis). Beyond this point, the texture mapping will no longer be squished, but part of the arm will be inside of the head (to compensate for its detachment), which can be seen by clipping inside of it.
 * 4,194,304 - 8,388,607 blocks

Piston arms now become extremely distorted after this point. Arms parallel to the direction of travel still look the same as their 4,194,304+ counterparts, but those which run perpendicularly are dramatically stretched to cover a full block's width. If on both axes, this can cause a vertically extended piston to appear like two stacked full blocks.
 * 8,388,608 - 16,777,215 blocks

Cauldrons and hoppers
First affected version: Beta 1.9 Prerelease 2 (cauldron), 13w01a (hopper) Last affected version: 1.7.10

Between these points, cauldrons and hoppers become very slightly distorted, in that the inner face becomes coplanar with the outer face. This causes the inside of the block to appear slightly larger.
 * 2,097,152 - 16,777,215 blocks

Cauldron and hopper inner face behavior becomes very interesting at this point onwards. Assuming one axis
 * 16,777,216 - 33,554,431 blocks


 * 4n: the internal faces are outright invisible
 * 4n+1: the inner face in the positive direction is missing
 * 4n+2: both inner faces will be offset outwards to adjacent blocks
 * 4n+3: the inner face in the positive direction is missing

Flower pots
First affected version: 12w34a Last affected version: 1.7.10

Beyond this point is where flower pots first become distorted. While not immediately obvious, paying attention to the dirt reveals that more of it is visible than previously. A flower pot at exactly 1,048,576 blocks will only be distorted on the positive half of the pot - the negative half will remain normal. Beyond this point, both the positive and negative halves on the applicable axes will be distorted.
 * 1,048,576 - 2,097,151 blocks

Distortion on flower pots becomes more obvious beyond this point. The internal faces of the pot now are rounded to be closer to the center, rather than slightly farther from it like before. As a result, looking down into the pot allows the sides to be seen out through.
 * 2,097,152 - 4,194,303 blocks

Similar to the previous entry, a pot at exactly 2,097,152 blocks will have the old distortion in the negative half, and the new distortion in the positive half.

It is worth noting for this entry and all subsequent flower pot entries that these internal faces are only rendered from the intended angle and are invisible from behind due to them never being normally seen from behind. This is intended design and is not a distance effect.

The flower pot's internal planes have now been completely ejected from the pot itself, such that there are now two full block pixels between the intended edge of the pot and the new position of the planes.
 * 4,194,304 - 8,388,607 blocks

Similar to the previous entry, a pot at exactly 4,194,304 blocks will have the old distortion in the negative half, and the new distortion in the positive half.

Interestingly, distortion beyond this point is now positionally dependent, despite the precision loss not being greater than a full block at this point. Assuming only one axis exceeds this coordinate:
 * 8,388,607 - 16,777,215 blocks


 * 2n: the internal planes simply appear absent. However, some faces now appear at full brightness where they would usually not. This may mean that the internal faces are offset to coincide with the pot's faces in these cases, as the internal faces are always drawn at full brightness in these versions.
 * 2n+1: pots will have their internal planes outright offset into adjacent blocks.

For mixing offsets, pots at double even coordinates have no inner faces and are fully bright, pots at an odd-even coordinate will have two fully bright faces and two offset faces, and double-odd pots will have all four faces offset (which can all be seen at once).

Distortion increases again beyond this point. Assuming only one axis:
 * 16,777,216 - 33,554,431 blocks


 * 4n: have no visible inside faces and some fully bright outside faces
 * 4n+1: both of the inside faces will be offset in the negative direction
 * 4n+2: have no visible inside faces and some fully bright outside faces
 * 4n+3: both of the inside faces will be offset in the positive direction

Mixing coordinates functions as follows: for pots at double even coordinates, internal faces will be absent and the pot will be completely unaffected by ambient occlusion. At an odd-even coordinate, two outside faces will be fully lit, two inner faces will be absent, and the other two inner faces will be offset to the nearest multiple of 4 block. At a double-odd coordinate, two faces will be offset to one of the two closest multiple of 4 blocks, and the other two will be offset to the other.

Flowing water and flowing lava
First affected version: Infdev 2010-06-16 Last affected version: Beta 1.7.3


 * 16,777,216 - 33,554,431 blocks

As a fluid block's "footprint" always stretches across an entire block face, the effects of precision loss here can only be seen once precision itself can no longer represent blocks (integers) individually.

Interestingly, despite often having angular features, these are not actually lost with the precision loss as one may expect, and are somewhat preserved. However, the angle becomes clearly shallower.

As with other distance effects, exceeding each integer power of two will double the error systematically, allowing for fluid sides as long as 128 blocks to be witnessed in semi-vanilla bounds.
 * 33,554,432+ blocks

Rails, powered rails and detector rails
First affected version: Infdev 2010-06-18 (rail), Beta 1.5 (detector rail, powered rail) Last affected version: Beta 1.7.3


 * 16,777,216 - 33,554,431 blocks

As a rail block's "footprint" always stretches across an entire block face, the effects of precision loss here can only be seen once precision itself can no longer represent blocks (integers) individually.

Beyond this point, a rail block will be stretched along the axes exceeded. Only half (a quarter if on both axes) of all rails placed will render, with the remaining half still being existent, but not visible.

Sloped rails will become stretched in much the same way as flat rails. If stretched in the direction of vertical travel, this will also result in the angle the rail makes with the terrain becoming shallower, as can be seen in a screenshot below.

Gears and ladders

 * 1,048,576 - 16,777,215 blocks

Fences, fence gates and walls
First affected version: Beta 1.9 Prerelease 2 (fences), 12w01a (fence gates), 12w34a (walls) Last affected version: 13w06a (fences), 15w38a (fence gates), 1.4.3 (walls)

At this point, the hitboxes for fences, gates and walls begin to break. Before this point, pushing into a fence in the positive direction would allow the player to go as far as .07500 blocks into the space the fence occupies; beyond this point, the limit becomes .20000 blocks. It is also possible to easily see inside of walls due to this.
 * 2,097,152 - 8,388,607 blocks


 * 16,777,216 - 33,554,431 blocks

Hitboxes beyond this point become very strange, and virtually non-existent in some cases. Assuming this distance is exceeded on only one axis and in the positive direction, a fence at a multiple of 4 coordinate will have an extremely thin hitbox at the negative side of the fence. A fence at a coordinate one more than a multiple of 4 will have no hitbox at all, one at an even, non-multiple-of-4 coordinate will again have a thin hitbox at the negative side, and one below a multiple of 4 will have its thin hitbox at the positive side. The image to the right showcases the solid faces approximately marked by signs.

Cakes and cacti
First affected version: Alpha v1.0.11 (cacti), Beta 1.2 (cakes) Last affected version: 15w38a (collision boxes), 15w49a (outline boxes) Fixing of cake outline box impossible to see due to 

After 1,048,576 blocks, the first breakdown of hitboxes can be witnessed for cakes and cacti. Previously, their hitboxes would perfectly fit their cuboidal shape, however any placed past this point would experience a minor distortion; the hitboxes would be stretched to a full block width on whatever axis this distance was surpassed on. This also affects their collision boxes - previously pressing against these on their negative-facing side would get a value of .76250, but after this point this value is now .70000, like a full block. Cacti can still cause damage with this expanded hitbox.
 * 1,048,576 - 2,097,151 blocks


 * 2,097,152 - 4,194,303 blocks


 * 4,194,304 - 8,388,607 blocks


 * 8,388,608 - 16,777,215 blocks

Beyond this point, the hitboxes of these blocks start to become quite strange. As a block's collision box cannot extend outside of the block space it occupies, it becomes very easy to outright clip through the blocks in several cases. On one axis, the hitbox will either be a plane or stretched to occupy two different blocks at once.
 * 16,777,215 - 33,554,431 blocks

Infdev-Beta terrain rendering offset/jitter
First affected version: Infdev 2010-06-24 Last affected version: Beta 1.7.3

Many effects are noticeable after traveling millions of blocks away from the center of the map. The first effect that becomes evident from 0624 Infdev to Beta 1.7.3 is the jumpy or stuttering movement of the map. In old versions of Minecraft, when the game renders the map around the player, there are two different versions that are generated, the rendering of the blocks themselves (i.e. entities, how much the block generates) and the visual aspect to the blocks (e.g. the hitbox, block pixels etc.). In Java Edition Beta 1.7 and prior, the visual aspect of world rendering loses precision, resulting in the jumpy movement. The reason the world loses precision in the first place is that the rendering engine uses double-double precision, which determines the position of the world around the player. The rendering engine then uses a floating point integer to determine how far the already rendered blocks move around the player when they move. Unfortunately, this floating point integer cuts the double-double precision number in half, and thus the rendering engine doesn't have enough information to render the terrain correctly, and the information gets cut in half every time the player doubles their distance. Removing the floating text from the rendering engine fixes the stuttering movement.

This jumpy movement is notable even at an X/Z of ±16,384, becoming increasingly noticeable around ±524,288. The intensity of such glitches doubles every time the player passes a coordinate that is a power of two (e.g. 2,097,152 or 4,194,304). After about X/Z 16,777,216, the hitbox is a full block off, which makes placing and destroying blocks almost impossible. The floating point precision errors are no longer noticeable past X/Z 2,147,483,519, as surface textures stop rendering and blocks no longer generate. At the center of the world, the world would be off by only $1/4194304$ (0.000000238...) blocks. Past the 32-bit limit, the world would be off by 256 blocks. For any position between n and 2n, where n is a power of 2, the world is offset by $n/8388608$ blocks, or $n/524288$ block pixels.

At excessive X/Z positions (depends on the operating system and version, but usually occurs between X/Z: ±268,435,456 and ±2,147,483,647, and early versions such as Beta), world render glitches out, resulting in terrain that appears to flicker. This does happen at X/Z: <±268,435,456, but it becomes very noticeable at X/Z: >±268,435,456 and gets worse every power of 2 beyond that. This could most likely be that past X/Z: ±268,435,456 the Minecraft world is 32 blocks off, which is as large as 2 chunks, though it is unconfirmed.

Piston offset/jitter
While this bug did affect Beta 1.7, it was impossible to notice due to the rest of the world being broken in the same way, such that the two effects would occur in unison. While world rendering was fixed in Beta 1.8, piston tile entity rendering was not, resulting in pistons looking out of place as a result until their fixing.

Infdev 0227 chunk detaching issue
May have technically occurred in earlier game versions, but the world stopped rendering after 2560 blocks out anyway, so it would never have manifested anyway A similar glitch occurred in snapshot 14w30a/b.
 * X/Z ±131,072: Chunks begin to shake. This effect doubles for every power of two that the player walked away from the spawn point.
 * X/Z: ±2,097,152: World stops rendering completely at certain angles.

14w30a/b chunk detaching issue
Interestingly, this bug appears to only manifest with certain hardware. It has been noted with Intel integrated graphics. This bug can also happen in first person. This bug is fixed with VBOs.

Indev/Infdev hitbox rendering issue
The shape of the hitbox distorting and becoming non-cubic is also possible on some GPUs.
 * Every power of 2 that the player goes, the hitbox of the block that the player is facing becomes more and more distorted.
 * The hitbox becomes increasingly more corrupted and distorted until it disappears entirely at X/Z ±2,147,483,648.
 * X/Z ±2,048: Hitbox begins to subtly lose its shape.

Falling block rendering issue
The shape of the block distorting and becoming non-cubic is also possible on some GPUs.

Indev/Infdev entity movement

 * Every power of 2 that the player goes, it becomes harder and harder to walk along the axis the player is traveling on.
 * At X/Z ±4,194,304, the player falls through the blocks. However, if the player teleports beyond this point they can still stand on the blocks but cannot move along that axis.
 * At X/Z ±8,388,608, the player can no longer stand on the blocks and falls into the void.
 * X/Z ±16,777,216: Blocks are no longer solid; the player falls and hits a layer of lava.

Mods and map editors
Up until 1.1, there existed a major bug with the OptiFine mod  which would cause world rendering to completely break when traveling a far enough distance away, with effects manifesting as soon as 10,000 blocks. The effects would appear to get more intense continuously when moving far away from the center of the world, as opposed to having cutoffs every 2 blocks, due to this rendering bug also taking rotation into account (although if the rotation is kept constant and along one axis discrete jumps become noticeable). The bug existed due to the mod undoing vanilla's fix for said bug.