I have been experiencing the same issues with a major performance hit in 1.14.x. I created a new world 1.14.2-PR2, Java 1.8.0_74 64-bit on a MacBook Pro (i9-8950HK hexacore, Vega Pro 20 GPU, macOS 10.14.5). Upon creating this survival single player world with no options/flags changed, I moved in the negative x direction solely until I experienced the issue. Upon experiencing the issue, I opened my single player world to LAN w/ cheats enabled so I could access the debug profiler, and ran it for 3 minutes whilst staying completely still and not touching my computer. Afterwards, I also collected a sample of the Java process from macOS' Activity Monitor.
Symptom: After walking for about 1-2 minutes, every 3 seconds, the game freezes for half a second, with a visible TPS and ms lag spike being shown. Please note that the issue continues to occur in Peaceful mode. Please also note that disconnecting and reconnecting makes the spikes stop.
I realized I could be more useful by isolating more variables in my debug profile. So, I am attaching two debug profiles, of the same world, with me in the same place, doing the same thing (literally nothing-- not touching my computer). One was taken during this bug occurring after walking around the world for a minute, and the other was after I closed and reopened this single-player world (didn't move at all, server performing as expected). Peaceful mode set in both.
I hope that seeing how these profiles differ in the same spot in the same world in the same difficulty with the same actions taken during the debug profile process will help in isolating the issue.
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.
Hi all,
I have been experiencing the same issues with a major performance hit in 1.14.x. I created a new world 1.14.2-PR2, Java 1.8.0_74 64-bit on a MacBook Pro (i9-8950HK hexacore, Vega Pro 20 GPU, macOS 10.14.5). Upon creating this survival single player world with no options/flags changed, I moved in the negative x direction solely until I experienced the issue. Upon experiencing the issue, I opened my single player world to LAN w/ cheats enabled so I could access the debug profiler, and ran it for 3 minutes whilst staying completely still and not touching my computer. Afterwards, I also collected a sample of the Java process from macOS' Activity Monitor.
Symptom: After walking for about 1-2 minutes, every 3 seconds, the game freezes for half a second, with a visible TPS and ms lag spike being shown. Please note that the issue continues to occur in Peaceful mode. Please also note that disconnecting and reconnecting makes the spikes stop.
/debug output: https://i.succ.in/2xa8gcHD.txt
Sample of java process: https://i.succ.in/2x653Ohg.txt
Let me know if I can be of help in any other ways.
Edit: Issue persists after /kill @e[type=!minecraft:player]
I realized I could be more useful by isolating more variables in my debug profile. So, I am attaching two debug profiles, of the same world, with me in the same place, doing the same thing (literally nothing-- not touching my computer). One was taken during this bug occurring after walking around the world for a minute, and the other was after I closed and reopened this single-player world (didn't move at all, server performing as expected). Peaceful mode set in both.
Bug occurring (BAD performance): https://i.succ.in/2xm2t9eF.txt
Bug NOT occurring (GOOD performance): https://i.succ.in/2xrioEpF.txt
I hope that seeing how these profiles differ in the same spot in the same world in the same difficulty with the same actions taken during the debug profile process will help in isolating the issue.
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.