Bad server side performance since 1.14

Not necessary, it's in fact MC-132073

Now that MC-132073 has been fixed (some time ago), please check whether or not this issue persists in pre5.

Could not reproduce it with the current development version. Please send me feedback if this is still an issue in the next released version.

1.13-pre7. Running 1.12.2 world. Server can't even maintain 13 to 15 tps. (1.12.2 runs at 20 with room to spare). Chunk loading/conversion can take more than 60 ticks and trigger watchdog. 

Also still experiencing this on pre7, if anything the problem is worse. Converting a 1.12 world, conversion time took several minutes, took about 5 minutes before the game managed to drop down from 2000-3000ms to 15-20ms tps. After waiting for that to calm down, I still can't Elytra fly across our spawn area without freezes, fps drops, chunks not loading towards the edge, and eventually a crash. This has crippled performance of our server world to the point that it's unplayable now. @unknown @unknown 

@unknown  @unknown @unknown This issue has not been resolved at all. Why is it marked as resolved? My world is still unplayable. I have it uploaded to a Google Drive folder – if you send me an email to share it with, I can add you as a collaborator to download the world. 

Again, this issue is not resolved, please do not mark it as such. This is still a game breaking update for server owners and single player owners alike. 

Problem seems to stem mainly from chunk conversion, which appears to be done on-the-fly when 1.12.2 chunks are being loaded. That process is painfully slow. Some heavily populated chunks can take forever to convert. Worlds should be 100% converted at world load, no matter the time it takes. Otherwise, server owners will have to deal with performance issues and multiple crashes.

The only way I can actually explore my 1.12.2 world is by advancing on a chunk-by-chunk basis. Rubberbanding is the norm. There are areas where I can sneak faster than chunks can convert and load. And that's me, all alone by myself on my server. Now if we were 5 or 10, the server simply wouldn't run at all.

Once the chunks are converted to 1.13, performance goes up, but nowhere near normal operation. I've been standing still in a "not too busy" area of my world for 20 minutes, and still got bombarded with "Can't keep up!" messages in the console. All of them well over 60 ticks each.

After a little more testing I can echo @unknown's comments and findings. As a server admin the way things are currently handled is going to cause absolute chaos for our server, we cannot explain to players that massive dips in performance might be because people are reloading 1.12 areas in 1.13 for the first time since the update, and even then it might just be the generally worse performance. Do we just need to wait/hope for a mod to find this and mark it as 'not fixed'? Can't even get other community members voting for it while it's marked as fixed!

@unknown – Is this initial chunk conversion lag-inducing only for the initial conversion for anyone, or for everyone? Meaning, if I download my (large) server world, fly all around it and deal with the lag for hours on end while the chunks are updated in 1.13, then re-upload it, will that prevent other people from getting the same lag? There has to be a better way for Mojang to handle this upgrade. 

@Jfarf_Gl, it's global. If you fly around converting all the chunks and re-upload that converted world, it should be fine for everyone else.

Worth noting that as @unknown and I seem to be experiencing, even after this conversion is done in areas, performance still isn't close to what 1.12.2 offers in the same areas, and I at least am getting frame drops and stutters in some areas where issues never existed before.

Same here, even after initial chunk population and a very generous amount of time standing in the same chunks, I'm still not seeing the same performance as in 1.12.2, even not near oceans. 1.13 Pre Release 7. Again, I can send my world file, it's already uploaded into a Google Drive folder, just say the word and I will add an email as a collaborator. 

Is it much too late to get a utility to just convert the world before starting play? The conversion process is causing endless issues for me as well (eg https://bugs.mojang.com/browse/MC-132768), I'd much rather run something once as a process for updating, even if it takes hours (4+Gig of worl)

Still an issue in 1.13 pre-8.

Also confirming issue for 1.13-pre8.

Can reproduce this issue on the latest 1.13-pre8 snapshot

I have my world uploaded, I just need an email to share it with.

I also deeply want a standalone tool to convert worlds, or the option to do the conversion before the world is loaded. This could become a security issue in the full release: all players would have to do to crash small servers is fly with elytra and load too many 1.12 chunks at once until the server ticks too slow and crashes.

Confirming for 1.13-pre9. Although there's a big improvement over previous pre-releases, there's still a lot of rubberbanding and "Can't keep up!" messages in the console.

1.12.2 world have a bit better than the original snapshot, but there are still a lot of unbearable lag spikes 😞

Confirming for 1.13-pre10. If anything, it's even worse than pre-9, according to my debug profiler log (see attachment). World has even gone through an "Optimize World" process before testing.

I'm going to try out this tool tonight and will report back. I have a pretty large world, so it may take a while!

@@unknown

Depending on how large your world is, allocate tons of RAM to the JVM before clicking that button. My 700 MB world required 4 GB allocated to it, or it would crash from an out of memory error (MC-133808).
 

Update – World is about 530,000 chunks. I allocated 5 Gigabytes of RAM and it crashed about half an hour in, at about 55% progress. Trying it with 12 gigabytes allocated, but...this doesn't look like it's going to be the one-all-be-all solution. 

@Jfarf_Gl

While it may not be the ideal solution, it's still a great tool. Especially for large worlds on bigger servers. Shedding away the need to convert chunks on the fly.

It's most likely a serious memory leak. Unfortunately, my perfectly legit bug report has been wrongly resolved by some bot marking it as a duplicate of an unrelated 5 years old bug. And mods have been ignoring my comment to reopen the report. They've been active everywhere else, though. MC-133808.

Also confirming for pre10.

After allocating a whopping twelve gigs of RAM, world conversion worked (never seemed to go above six gigabytes, for anyone concerned). 

I will test performance tonight and report back. World conversion ('optimization') definitely worked in terms of completing all of the chunks, as the second time I ran it the same 530,000 chunks only took about 10 minutes. As for if this actually reduces lag/increases TPS, I will report back on that tonight. 

Anyone have an easy way to check for TPS without using a plugin? 

@Jfarf_Gl

Use the built-in debug profiler:

/debug start

Let it run for a little while, then stop it:

/debug stop

The look for the profiler log in the world's debug subfolder. You won't see your tps in real-time, but you'll get an idea of the average tps you got during the profiling.

It's not perfect, but the F3 menu now shows the ms taken to complete a tick, you can extrapolatie it from that. EDIT: I had no idea about /debug start, Early Reflections suggestion is miles better.

Can you confirm whether the world conversion/optimization also converted the Nether/End? Nether should be easiest to confirm through performance. 

@DaUltraMarine

I can't confirm for sure, but after doing the optimization, my DIM-1 subfolder (Nether) grew in size. However, my DIM1 subfolder (The End) remained byte-for-byte identical.

My best guess is that converting The End is useless, as region files and chunks have very little data compared to the other dimensions. They're already lightning fast to load.

Okay everyone, after running the optimization, I can confirm that, for my world at least, performance is heaps and bounds better. It's a night and day difference before and after the optimization. I did get a lot of errors in the log when I flew around the world (a few oceans, some extreme hill biomes, some plains), now attached as "latest-after-optimization.txt". I did notice a bit of stuttering here and there, and again quite a bit of errors in the log, but I can say that my world will definitely be more than playable in 1.13 now. 

Edit – after playing around a bit more, world is still quite playable and much better than before, but there are some bad lag spikes as mentioned by @@unknown 

@Jfarf_Gl

I did manage to get some more tps by doing some cleaning in my world. According to my own debug profiler log, I saw that hopper minecarts took up quite a bit of ticking. My average tps got up to 18.87, which is pretty good in the circumstances. But that was me standing still for 2 minutes. Just flying around got it down again to 17.

I also see that villagers take up most of the ticking. I didn't bother getting rid of them. I know I would have gotten back most of the missing ticks.

But still, the point is I can't manage to get the server up to ~20 tps without #1, getting rid of all the hopper minecarts in one of my contraption and #2, getting rid of most of my villagers. Now, 1.12.2 can handle all of those at ~20 tps without any hiccups. (when I say ~20 tps, I mean around 19.95, as per profiler logs).

My conclusion is that entities (mostly mobs and animals) gobble up more ticking resources than before. I may be wrong though, but that's what I see.

So, from my point of view, this report still stands for bad overall performance. But is it due to being a 1.12.2 world? I don't think so. I suggest you update the report to "confirmed for 1.13-pre10" but change the title by removing the 1.12.2 world reference. It's just overall poor performance.

Update - in 1.13 I'm still averaging only about 16TPS in most areas, with the occasional bad lag spikes here and there. Can post more information as needed (and, again, a world download that nobody has provided me an email to send to). 

Another update - I re-uploaded the "optimized" world to my Java Realm and....it's not good at all. Incredibly laggy, nobody can even use an elytra now, mobs will be on fire for 3-4 minutes before dying because we are going as low as ONE tick per second...it's not good at all. Performance is much worse in multiplayer, borderline unplayable. I'm incredibly frustrated since I'm on a Realm so reverting isn't possible.

I moved my spawn chunks to the middle of nowhere which did help with the lag slightly, but that's a pretty awful workaround because (a) it's still really laggy and (b) all of our farms are in spawn chunks.

We only had two people on at a time yesterday....I can't imagine if we tried having eight people on the Realm at once. It probably would have crashed.

Can confirm this for a brand new 1.13.0 world. Chunks load very sporadically. I also get block lag and some blocks just won't render their texture at all( I can see the hitboxes outlined). This happens both on loading and generating chunks. I have no performance issues on 1.12.2 with vanilla mc( no optifine). One thing worth mentioning is that performance improved a lot when I limited my fps to 30 via the in-game option, so this might be some timing issue between threads and/or draw calls. I'm using linux( fedora 27) with a dedicated amd gpu and a quad core i7 that doesn't throttle so performance should be plenty.

Lots of people on Reddit posting about poor performance in 1.13 worlds both old and new. Looks like 1.12.2 worlds were hit hard in particular, but brand new 1.13 worlds are seeing lower performance than they should be as well. 

Edit – Best example I can use right now: I have a command block set (probably 70 or so blocks) to remove all of the snow from my castle area. The commands all follow this format (updated for 1.13):

/fill 350 91 1550 400 102 1600 minecraft:air replace minecraft:snow

I hit one button and all 70 or so command blocks turn on and remove all of the snow in my area. In 1.12.2, this happened seamlessly, for almost an entire extreme hills biome. Absolutely no stuttering while the commands are being run. In 1.13, running these command blocks one time will drag the entire server to a halt. Again, there could be 8 people on the 1.12.2 server, all in different areas, while I run that command and it has literally zero noticeable impact on performance. On my 1.13 Realm, same exact world, I can be the only one online and the same exact commands being run grinds the server to a halt. 

@Jfarf_Gl

This is likely a direct effect of the flattening. Everything in 1.13 has a unique string ID. Before 1.13, all the ids were integers, internally. It's much slower to process and parse strings than dealing with integers, which is the fastest possible route.

Also, dealing with strings all over the place is both more demanding on CPU and memory. That's why I once said that some aspects of 1.13 are inherently slower and more demanding "by design". All the imaginable optimizations will never bring back 1.12.2 performance levels in a lot of cases.

You can do a search for "Clean code vs efficient code". Clean code is preferable because it makes coding easier and more readable. It's also a lot easier to maintain and expand upon it. It's more "future proof". But it's also a lot harder to make it really efficient (fast).

Also, clean code means "no more shortcuts" and "hacks". Those are questionable because they can be unstable and unreliable. Updates can break them. But they also (usually) provide a much faster route to a result (like bypassing an API to write directly to the memory or the frame buffer).

My best guess is that the old Minecraft codebase made use of shortcuts and hacks so much that it became a mess (spaghetti code) that had to be dealt with ASAP or it would have become unmanageable.

@EarlyReflections   Thanks for that informational write-up! I can deal with the lag in the case of that snow removal command being run as it's very situational and doesn't last very long.

The issue is that there is still high levels of lag in general, and chunk loading seems to be very bad. Everybody on my server has reported lots of issues with chunks not loading in (sometimes I'll have chunks in front of me NOT load in, but other chunks behind those ones, farther from me, load).  Trying to fly around is impossible; the only way we've been able to play in 1.13 is by having everyone explore the same chunks together, and even then there's still lots of lag. 

There has to be more Mojang can do to optimize this update. 

@Jfarf_Gl

Yes, chunk loading performance is definitely an issue. Here's a comment from Amidst developer describing how 1.13 chunk loading is actually up to 35x slower than before:

https://github.com/toolbox4minecraft/amidst/issues/395#issuecomment-386838958

The whole thread is worth reading. This was a while ago and things probably got a bit better since, but nowhere near 1.12.2 levels.

I can't remember exactly when (probably in the 1.8 development phase), but Dinnerbone once pulled off a little miracle by threading chunk loading code, making them load almost 10x faster. Well, it's time someone pulls off another one.

Similar issues for me. It seems to be especially laggy around redstone circuits - even those of the simplest design - and beacons. Reducing graphical settings changes nothing.

I've got a Ryzen 1700X and a GTX 1080, with updated 64 bit java, the latest chipset and graphics drivers, and freshly TIMed components. There's no hardware issues of which to speak.

For the record, I play on a multiplayer server that I host. All of my users are experiencing the same issues, and the server logs the same error message periodically, claiming it "can't keep up" and asking if the "server is overloaded". No amount of fiddling with JVM arguments makes a difference.

@Jeff Gaunt

— It seems to be especially laggy around redstone circuits

Yes. I've read somewhere (can't find source anymore, sorry) that redstone dust, which was already screaming for fixes because of lag (badly implemented floodfills), is a lot worse in 1.13. We need someone to actually confirm that. 

I finally got Optimize World to work and my previously unplayable local copy is now playable.  Dinnerbone has said that it "secretly" is updating chunks in the background to the new version.  I'd say it's not all that secret.

I'm trying to figure out if I should upload back to the linux server, my Optimized World MCAs or just update the server and leave it alone for a while.

I run a server for a myself and a few friends, and am having this issue. We started our world in 1.12.2 and stuck mostly to the Nether, then with the 1.13 release replaced a new world's Nether with the 1.12.2 one. Our new 1.13 Overworld is completely fresh, and a little less laggy than our developed Nether bases, but that may just be because of all the redstone etc. that our Nether bases have, I'm not sure.

I tried booting up a fresh 1.13 server for us and still experienced lag, though, including that impossible rubberbanding elytra flight, and random chunks just refusing to load.

It's not simply annoying. The lag makes fighting incredibly difficult, Ghasts are blowing up all my builds, etc. I can't play it like this.

Frame rates/freezes are worse than in the snapshots. No clue why, I am using the same world converted From FROM the snapshots... aside from my dirt hut and five double chests of crap, there is nothing to this virgin world! What happens when I load fifty chests and create my standard hopper/sorting room inside my base? If this is a problem with NO constructions, no redstone running/present, no timers etc... then I cannot rely on this version to run a damn thing later! Why play a world for two months to get to a point where the damn thing crashes?

I have worlds playable with far more farms/timers/redstone/updates ongoing, and still they perform better than this version with NOTHING in it??

If I am impressing you with an urgency to fix whatever is wrong with 1.13, so much the better!

I'm having the same issue. A simple redstone repeater clock kills the cpu usage and causes the server to be consistently 250-300 tick behind. Crashes are common.

Confirming for 18w30a.

Optimizing the world helped my server a ton. I can actually go to to my mob farm and the server only runs 40 ticks behind instead of 300. But I'm still getting way more lag then I should.

My guess is optimization isn't in their top priority list right now, so many bugs need to be squashed down, I feel like the 1.13 release is still in a snapshot state and shouldn't have been released yet, but before release, there's not that much about bug reports, you start seeing them once something is officially released, cause I'm going to say perhaps only 15-20% of Minecraft players test the snapshots, the others wait for real release.

Confirming for 18w30b.

Confirming poor optimization and (arguably worse?) chunk loading for 18w30b. I just attached a screenshot of another example of the poor chunk loading – there has to be less than 30 blocks in this chunk yet it refuses to load until I disconnect and reconnect multiple times. Screenshot is of a chunk in the End.

TPS is still very low as well, nowhere near 20. 

Still waiting for a resolution. My server keeps coming back with the following repeated errors:

[22:17:55 WARN]: Can't keep up! Is the server overloaded? Running 2372ms or 47 ticks behind
[22:18:12 WARN]: Can't keep up! Is the server overloaded? Running 2039ms or 40 ticks behind
[22:18:49 WARN]: Can't keep up! Is the server overloaded? Running 2001ms or 40 ticks behind
[22:19:15 WARN]: Can't keep up! Is the server overloaded? Running 2003ms or 40 ticks behind
[22:19:43 WARN]: Can't keep up! Is the server overloaded? Running 2007ms or 40 ticks behind
[22:20:02 WARN]: Can't keep up! Is the server overloaded? Running 2001ms or 40 ticks behind
[22:20:21 WARN]: Can't keep up! Is the server overloaded? Running 2003ms or 40 ticks behind
[22:20:39 WARN]: Can't keep up! Is the server overloaded? Running 2604ms or 52 ticks behind

No matter what I do to the JVM arguments or server settings, the results are the same. It seems that every time I face a large group of passive mobs, go near a beacon, or look at a redstone circuiit, things stutter like crazy. Any word on a solution yet?

Getting the same issues here. Breaking log blocks from trees is also a huge performance issue for some reason. I have used the "optimise world" option before I played on my old world in 1.13.

New 1.13 worlds also have the exact same issue.

I too am having the same problem.

Ubuntu VM running on a AM3+ FX 6300 Host. The VM has 2 cores allocated and 3GB ram. We mainly play locally over LAN

Its been running for several years now, running various versions as they come out. When we moved to 1.13, we created a new world, and then the lag started.

We see classic "rubber banding" lag, and the server states it is lagging behind. 

I increased the VM allocation of virtual cores to all 6, it has had some effect but it is still very laggy. The server even stopped at one point today as the lag exceeded 60 seconds!

We've only had 3 players in this server at any one time so the specs should be perfectly suitable, but its not.

[14:17:15] [Server thread/WARN]: Can't keep up! Is the server overloaded? Running 12912ms or 258 ticks behind
[14:17:39] [Server thread/WARN]: Can't keep up! Is the server overloaded? Running 8954ms or 179 ticks behind
[14:17:59] [Server thread/WARN]: Can't keep up! Is the server overloaded? Running 5681ms or 113 ticks behind
[14:20:30] [Server thread/WARN]: Can't keep up! Is the server overloaded? Running 2006ms or 40 ticks behind
[14:21:00] [Server thread/WARN]: Can't keep up! Is the server overloaded? Running 15105ms or 302 ticks behind
[14:21:30] [Server thread/WARN]: Can't keep up! Is the server overloaded? Running 14950ms or 299 ticks behind
[14:21:55] [Server thread/WARN]: Can't keep up! Is the server overloaded? Running 10548ms or 210 ticks behind
[14:22:16] [Server thread/WARN]: Can't keep up! Is the server overloaded? Running 5708ms or 114 ticks behind
[14:22:35] [Server thread/WARN]: Can't keep up! Is the server overloaded? Running 4106ms or 82 ticks behind

Confirming for 18w31a. Things got better. I do get more encouraging numbers. It's not there yet, but it's on the right track. 

@sewan - we had to increase the max tick time to 10 minutes to stop crashes. The closest thing we can find to a pattern is that the conversion on some old chunks gets into a weird state sometimes but does eventually sort itself out? Not a great help

I have to say that the 18w31a lag is not as frequent, but if it is stuck, it is still very serious and even crash.

Lag seems to be better in 18w31a, but chunk loading is still in a very bad state right now. 

I converted one of my redstone test worlds to 1.13 to update some of my designs. I had a mob farm that uses observers with string to detect mob spawns then push them to their death with pistons. The observers were stuck in the on position and I had to replace every single one to get the farm functioning again to test it. Removing the string in front didn't update them out of their bugged state. After it started working I was getting terrible frame drops I normally sit around 100 FPS and it was dipping down to 15 FPS at times. In 1.12 this design didn't cause any frame drop/loss issues in 1.13 it's pretty noticeable. I haven't tried recreating the farm in a new world yet to see if it still causes FPS loss.

Playing stable Minecraft v1.13 with minimal graphics on MacBook Pro (Retina, 13-inch, Early 2015) macOS v10.13.6 (17G65) even discharges my battery when it's 100% and charging. That happens when server is running on another PC.

I fixed my performance issue (Bug 134611) by limiting my FPS to 30-40 FPS. It happens when the CPU can't catch up with the stuff that's going on in the game. When that starts happening the ticks become unplayable and the CPU becomes unable to catch up. By limiting your FPS to 30 or 40 (30 is recommended) then you'll notice less performance drops and the ticks will be averaging around 10-30 meaning that your CPU is able to catch up with the stuff in-game.

My issue was mostly mitigated by setting fullscreen resolution to 1152x720@60fps. It allowed me to restore better graphics.

Confirming for 1.13.1-pre1.

Some stuff definitely got better, such as chunk creation/loading. But something really bad also happened that hits the mspt really hard (probably entity collision management). I can't even make it go below 80, just by being around my base, but the average being around ~100. I could sometimes get ~60 in previous snapshots, which was already bad enough.

As reference, my average mspt in 1.12.2 is ~35, standing at the same exact place.

The problem

There are big performance issues with versions 1.13 and above of Minecraft.
A lot of lag spikes appear when the game loads and unloads chunks.
Also, when you open the pause menu of the game, a huge lag spike occurs.

Here are some videos of the problem...

A – When loading new chunks

B – When loading already generated chunks

C – When you open the pause menu

How to reproduce

1. Create or load a world in 1.13 or above
2. Once in play, press Alt + F3 to display the lagometer graph
3. Explore your world and keep an eye on the graph

Also, you can compare the results you get in 1.13 with those you get in 1.12.2. You will see the big difference !

I've decided to stop playing Minecraft 1.13, I'll wait till this is really fixed, its ridiculous how this game was released so badly optimized and now any simple redstone contraption for mob spawners or automation makes the game lag like crazy. Its an awesome update and I love all the new changes and the Aquatic Update, but this Vanilla update is worst in performance and fps wise than when I play Foolcraft 3 modpack which rely on a crap ton of mods and is on 1.12.2. So its clearly not my PC that is the problem. And its not only redstone, but just roaming around makes the game lag like crazy. RIP Aquatic Update, lets hope Mojang really cares about optimization soon enough. Oh yeah, its a brand new world created in 1.13 so its not an old world conversion problem either. At first my adventures were nearly smooth as butter, but the more stuff I build, the worst the performance become. Also I tested to run the server on a better PC and on my Main rig, it doesn't change anything, perhaps a little better at loading chunks on my 2nd rig that has a better SSD. 2nd rig is I5 3550, 16GB, GTX 770, Windows 8.1 Pro 64 bits.

Main PC Specs: 

Intel I7 4790K stock 4.0GHZ speed

16GB DDR3

GTX 770 OC WF3 4GB edition

Samsung 840 Pro SSD 512GB edition

Windows 10 Pro 64 bits

Latest version of Java 8

Server running the world Vanilla in 1.13 pre1

I'm the only player on this world since its family only.

Intel Q9400 2.6GHZ

8GB DDR2

Samsung 840 Pro 256GB

Video card irrelevant since its a server

Wndows 7 Ultimate 64 bits

Latest version of Java 8

Tickrate is much better in 1.13.1, however it's still not close to where it used to be, and chunk loading is still abysmal. Attaching more screenshots of the chunk loading issues I'm having – currently I can't even get half of my mountain to load in. 

Confirming for 1.13.1, even though it's light years better than in 1.13. Redstone is still begging for optimization, especially Redstone Dust updates. They were already painful before, but since the flattening, they are agonizingly sluggish.

@@unknown 

It seems like your issue isn't exactly with chunk loading but rather chunk rendering, client-side. The chunks actually load correctly, but take forever to render, or never do unless you walk close or right into them in order to see them. When you fly with Elytra you can see them appear in a more or less random checkerboard pattern. Chunk rendering is still very clunky. Although it isn't affecting speed directly, it still falls into the "bad performance" category, for sure.

It's better but still present in 1.13.1 release

I've found giving my server more memory has helped a lot. More of a workaround than a fix, but at least its playable.

Edit: Its a VM with two 3.52Ghz cores, and has 5Gb RAM assigned. This seems to support about 2-3 players before we see any noticeable lag

I start the server with the following switches:

-Xmx3G -Xms1024M

I can partially confirm this for 1.13.1 for a few actions, most notably breaking blocks around lots of leaves (e.g. cutting down trees), Here's an F3 screen while the spikes occur (the same happens in the bundled Java): https://i.imgur.com/lnEp8BG.png

Experiencing horrible tps issues here as well.  New world as of 1.13.  When 2 people are on the server, no problem.  As more people join, tps drops.  We currently have 7 people on, tps is at 10.  On 1.12.2, 7 people wouldn't hit tps at all.  We tried a different node on our provider using a backup server, tps was the same on a new 1.13.1 world as our primary 1.13 world.  Switched our secondary server to 1.12.2, same seed, everyone spread out to their base locations, tps is fine, even with the world generating.

I can confirm this. Very high lag spikes in the latest version 1.13.1.

You can find a F3 screen comparison for 1.12/1.13/1.13.1 on https://imgur.com/a/PQIZwtw

The worlds were regenerated every time from the same seed and i used always the same game configuration.

This game is broken, my gpu usage  is 15% while in game. also found a bug that if you do /kill your perfomance wil drop so bad that my fps was about 20 to 12 fps after a restart of the world it was working poperly 

Cpu I5 4460 

Gpu Gtx 970 4gb asus strix oc 

ram 12 gb ddr3

These issues still persist in stable 1.13.1, on both old and new worlds.

Chunk generation and chunk updates in general create rather absurd, and unacceptable lag spikes.

This issue is new since 1.13, as before i had no trouble reaching 170 to 200 FPS on a 2014 built mid range laptop.

Since 1.13, that performance has dropped to 60-75 FPS under the same settings.

Turning off VBO's does not make much of a difference, other than in, obviously, CPU usage.

I understand the engine as a whole went through a lot from 1.12 to 1.13, but the current state of this build is just not okay...

CPU: i7-4700HQ @ 2.4GHz

GPU: GeForce 840M 2GB

RAM: 8GB (4GB*2) DDR3 @ 800MHz

I'm also experiencing performance issues issues with 1.13.1 on the server side.  I have a small Linode hosted server and the upgrade from 1.13 to 1.13.1 made a big impact.  We have a spawn area that seems to be the culprit.  There is a lot of redstone, water mechanics, and villagers in the spawn area, but it did not seem to be nearly as much of an issue in 1.13.  I've also run similar stuff before 1.13 and never had issues unless there were many people on the server.  I've upgraded the server now and will report if it is still an issue.  Was previously running on the 4G 2CPU Linode config with no issues until 1.13.1.

Same issues here in single player on a newly built world. Blocks disappearing / reappearing. Getting hurt by random invisible hazards while mining, mobs take strange paths and aggro weirdly (or not at all), combat is so laggy you never know if you actually hit something. It's pretty awful.

I can clearly tell that this bug is real, but I only get lag spikes and a little tick lag. With a server, It's that and lag back (The server teleports the player back to a previous spot because they moved to quickly over a period of ticks. This can happen normally if the server is lagging or if the player is lagging).

CPU: AMD FX-8370 Eight-Core Processor

GPU: NVIDIA Geforce GTX 970

Ram: 8 Gigabytes

Dedicated Ram: 1 Gigabyte for this test only. I normally play with 4 dedicated.

If you haven't already, use the following JVM arguments for all profiles with 1.13 or above. This is the new default as of the update and makes the game run a little faster, but it doesn't get rid of the bug: -Xmx1G -XX:+UnlockExperimentalVMOptions -XX:+UseG1GC -XX:G1NewSizePercent=20 -XX:G1ReservePercent=20 -XX:MaxGCPauseMillis=50 -XX:G1HeapRegionSize=16M

To get these options an easier way, just make another profile in the launcher.

I uploaded a video showcasing how bad the lag is. Block updates are incredibly glitchy right now, and when a block is broken, the entire server lags (shown in video). 

I hope this is good place to put this: The prismarine animations could be causing the lag at the monuments in the ocean. Fish pathfinding can be causing the ocean lag. I have watched mobs (any kind) spawn and despawn 30 blocks away from me. Particles drop FPS quite a lot. (Entities have always causes lag in Minecraft.) A world I made in 1.13 starts out VERY laggy when loading up for some reason. After that it is alright. Can't remember anything else atm. Hope this helps, if not here, somewhere. 

Try the optimize world button for giggles. See if it helps any if you haven't tried it already. It might take its grand ol' time but I guess it would be worth trying at this point. I think it fixed my world.

Yes I'm also having unusually terrible performance in 1.13, it seems like nowadays the FPS goes down every update 😞
Huge lag when just walking around in single player: (video link here)

I don't have any issues at all with performance on single player, even when running tons of command blocks constantly, but on a server it's basically unplayable (survival mode with no active command blocks and minimal redstone mechanisms), with position resetting to an earlier spot, not being able to tell when you hit a mob or when they hit you, and blocks reappear after breaking them. It seemed to happen when the server switched from 1.13 to 1.13.1 (it also isn't my server, so I don't know any of the details about it).

Confirmed for 1.13.2-pre1.

Still in 1.13.2-pre2

I can confirm that fps instability while loading chunks is also apparent in 1.13.2.
I have video evidence here: https://youtu.be/mMejg24oHjA

I support this post because even with the latest update, the game is still unstable due to many lags spikes.
I am mapmaker and since our passage in 1.13.2, Elytra passages of our map have become unplayable.

Please try at least to have a performance level a minimum equal to the previous version 1.12.2! What is the usefulness of an update if the evolution of the game is a regression...

I can also confirm, in 1.13.2, with 6 server members and usually 2-3 online at a time, we are unable to use elytra, boats, or large numbers of shulker boxes without experiencing severe TPS lag. To the point that we get dropped out of the sky if we try flying around. Memory usage during these episodes is usually around 20% of the available 2GB.

I'm using the default JVM arguments in 18w47a, and the internal server (tps) stops randomly for over 20 seconds. The highest I've seen it so far is 47 seconds. During this, the frame rate can stop at much less, but still random intervals (100 fps normally), and can last for over 4 seconds.

Edit: I did this on an old world, but the previous snapshot barely had any lag in it.

Issues regarding performance with the 1.14 snapshots:
https://bugs.mojang.com/browse/MC-138114
https://bugs.mojang.com/browse/MC-138071
https://bugs.mojang.com/browse/MC-138676

there is bad performance also on server side constantly saying server behind thousands of ms.

1.14 snapshots confirmed

To fix this they should make blocks be integer storage or even compressed. Then also have a master map between id and block either in the world or have only a map per chunk of what the id means. This would also speed up clients as they receive smaller packet sizes.

They also broke something else the loading time shouldn't take this long to tick once chunks are loaded nor should the server be breaking.

Confirmed in 1.9 Bedrock
The items when broken by piston don't rise to the surface unless 3 blocks away from the item in realms and smp

This report does not encompass Bedrock Minecraft versions, as this is a Java Minecraft report. If you're experiencing issues, please create a report for Bedrock Minecraft here.

This is still happening to me on 19w14b. Ever since 1.13 the game has had extreme stutter while breaking and placing blocks. I am trying to make this piston elevator: https://www.youtube.com/watch?v=rTo5Ic0-Khk and every time I activate it the FPS drop to single digits. I have only made about 1/10 of the total height i'm aiming for. It's only 6 blocks tall right now. Even placing the blocks to build it causes extreme lag.

I activated hidden FPS and the spike goes halfway up the screen. I'm using 1440p with a GUI scale of 3 so the GUI is small. I have 8gb RAM allocated and am using 16 chunk render distance. My computer is not slow and gets ~500 FPS with uncapped FPS. I have a GTX 1080 and i7 7700k.

For placing blocks causing lag, try MC-123584. I typically stay in deserts or badlands so this won't happen.

Still in 1.14 Pre-Release 1

Still in 1.14 Pre-Release 2

So at 1.14 Pre-Release 2 not more

Hopefully it will be fixed before the 1.14 release 🙂

It is better but still in 1.14 pre-3, 1.14 pre-4 and 1.14 pre-5

Unfortunately, this problem has not been fixed from 1.13 to the current 1.14 pre5. Next Tuesday, 1.14 will be officially released. Should we wait until 1.15 to fix this problem?

Wow! Still in 1.14 Release! It's sooooo bad dude!
We have the impression that we came back in 2018 with the first snapshots of the 1.14 with the new rendering engine!

How was it done to regress at this point? It was not too bad in Pre-Releases...

Definitely still an issue.  I just updated to 1.14 from 1.13 and chunks load so much slower, I can outrun them, especially on a horse.  I literally have to stop and sit still for multiple seconds for the ground below me to load.  1.14 is full on unplayable in its currents state of chunk loading.  Good lord it's terrible, legit broken.

The loading of chunks is better but there are still a lot of lagspikes. So, still in 1.14.1 Release

Same, I can confirm that I am getting a lot of messages in the server console that the server is running 2,000+ ticks behind. Brand new vanilla 1.14.1 world. Lag spikes are causing a ton of rubber banding. 

The lag spikes are extremely bad in 1.14. When a player tries to connect to my server, they can't connect until a few attempts have been made. 

[ User Authenticator #2/INFO: UUID of player katona781 is 38918aa0-d0a9-4b8b-8876-31d99c662281
[Server thread/INFO]: com.mojang.authlib.GameProfile@214776fc[id=38918aa0-d0a9-4b8b-8876-31d99c662281,name=katona781,properties={textures=[com.mojang.authlib.properties.Property@1144c8ce]},legacy=false] (/68.41.40.37:61504) lost connection: Disconnected
[Server thread/WARN]: Can't keep up! Is the server overloaded? Running 193767ms or 3875 ticks behind
[User Authenticator #3/INFO: UUID of player katona781 is 38918aa0-d0a9-4b8b-8876-31d99c662281
[Server thread/INFO]: katona781[/68.41.40.37:61683] logged in with entity id 1149 at (-407.30000001192093, 73.0, 202.01249998807907)
[Server thread/INFO]: katona781 joined the game

Can someone please pick this ticket back up?

I can also confirm that this is still an issue.

Still effects 1.14.1
Happens when people teleport mostly via netherportals/commands

I agree, I've been running a Minecraft Server since 1.3.1. I was running 1.12 on a Xeon X5672 3.20GHz (4 Core) Poweredge R410, running on CentOS 7 and latest Java. It ran great. Then 1.13 came out, lots of server lag. Then 1.14 came out, more server lag. Lots of "Can't keep up! Is the server overloaded?" messages. I definitely mean NO disrespect to any of the programmers. I've done programming in the past, and understand how hard it can be to do. I think the next big update that should come as an "Optimize Only" update. Bring back 1.12's performance, please. I love the game, and have for years. Thank you for all of your hard work!

Still in 1.14.2 Pre-Release 1 and 1.14.2 Pre-Release 2

From my observations, I have found that erasing the lighting cache on startup is a temporary fix. But the bug will start again when the server stops sending chunks. My guess is that it's related to lighting changes in 1.14, which seem to be significant internally given the amount of other lighting bugs reported that appear at first glance to be unrelated to this.

The CPU usage in 1.14.2 is still incredibly high. Only one player in the game and the server suddenly used something over 1 GB of RAM and CPU usage was still too high.
When discovering the map and generating new chunks and also moving on already discovered chunks the performance was bad and the console was spamming information about player moving too quickly.

Here is the debug log of a clean vanilla 1.14.2 server and latest.log of the same run:

Run was done on Linux running Java 8u191 on 64-bit architecture.

As gamehosting side we still have to consider this version unoptimized as it is still too demanding, especially since the version 1.13 was perfectly stable and not resource-hungry as this new version. We still don't understand what happened.

1.14.2 still affected. Performance is abysmal, and getting into a nether portal is basically a guaranteed crash (as with all actions that force the server to generate more than ~2 chunks at a time). Meanwhile, the player-side rendering glitches into oblivion, as seen on the screenshot attached to the issue (namely, javaw_suuGljlf8N.png).

I'm very sad that this is still in 1.14.2. I really hope that the team at mojang is working on fixing this behind the scenes-- it's probably such a big issue that they're having a lot of trouble correcting it. At least, that's what I tell myself to sleep at night after seeing things like "wolves not sitting correctly on slime blocks" get fixed. 

Were you using mods?

Discussion for all reports on Mojira related to Performance Errors.

1.14.4 still affected, my private vanilla SMP with 4 GB of RAM that uses light datapacks only still rubberbands me every couple seconds, actually seems even worse than 1.14.3

Huge performance issues and crashing still on 19w35a
More than one player on and And crashing often, crashes every ~30 mins "stackoverflow"

Can confirm HUGE performance issues on multiplayer 1.14.4

Yes, still on 1.14.4

1.14.4 has already been marked as affected.

I want to note here that this ticket is not very helpful to the developers. It would be way more helpful if people could actually tell us what is causing lag in your world. Otherwise Mojang cannot do much about this. You can find out what exactly is causing the lag by using the /debug command.

As I expect that everyone here has their lag caused by something different, please create new and detailed tickets about your separate issues. A huge ticket about just general lag is not helpful at all.

Hi everyone! A bunch of fixes for server-side performance have already been done in the snapshot series for 1.15. I see that recent snapshot versions are still marked as affected. If you are able to reproduce these issues in the latest snapshot, please attach new profile information (/debug start and /debug stop when you have poor server performance). If possible, please share world files for affected worlds and details about where the problem occurs as well.

i can tell you, its trees. if ppl have massive tree farms the decaying leaves will freeze up the server, this bug doesnt appear to be happening in single player tho.

it also seems to happen when world save of overworld is greater than 10 gigs. any SMP earth thats pre-gen will do this. reminder its not a crash that happens its a deadlock.

A major issue with the lag is the way many blocks are ticked. Fire, trees, etc all have huge effects on performance in mc which completely hammer the CPU, or most likely, one core of the CPU.

wander why mojang cant enable multi core on server and client, every single pc on the planet now is multi core. why cant they upgrade that?

Heres a simple SMP earth map that simple server cant run
mediafire.com/file/ci93do7wrwh51ev/Earth.zip/file

also added comment, the deadlock will not show ANY error/logs even when using /debug commands. the bug happens when loading/unloading chunks rapidly AKA going from overworld -> nether -> overworld. the deadlock means theres an infinite looping code running that cannot be broke. from what i heard is that every chunk has its own block ids, this feature is whats causing the deadlocks.

Also with modding everyone uses java 8 update 221, so make sure u check this out with that version of java.

UPDATE
so after doing some testing i found that server cannot sync with player data from client for some reason which leads to a massive deadlock when a player joins/dies-respawns/changes dim, this is in 1.15.1

@@unknown Look at MC-188163 and MC-190258.

Can confirm.

can confirm 1.16.4 rc-1