-
Notifications
You must be signed in to change notification settings - Fork 22
Configuration~ Fabric 1.14 and 1.15
Thread count changes the number of threads used by the workers. Workers or executors are used for running tasks in parallel using multiple threads. The general rule is that increasing the values will increase CPU usage and may improve performance.
Technical detail: this modifies parallelism
in the ForkJoinPool
constructor.
Option | Smooth Boot Default Value | Vanilla Default Value | Notes |
---|---|---|---|
Main Worker Thread Count | 1/2 of total CPU threads | Total CPU threads, up to 7 | Main Worker is used by most parallel tasks such as stitching sprite atlas. Change this affects loading times, and may have an impact on in-game performance. |
Thread priority changes the priority given to each of the threads. The parameter is between 1 and 10, with 5 being the default. Changing this is similar to changing an application's priority in the task manager. A lower value leaves more performance for other programs, and a higher value increases Minecraft's CPU priority. Counterintuitively, it's a good idea to lower the thread priorities, at least for the worker threads. This stops Minecraft from lagging other programs that are running at the same time (ex. Youtube videos), while it does not have a large impact on the actual game performance.
Technical detail: this calls Thread.setPriority
for each thread when initialized.
Option | Smooth Boot Default Value | Vanilla Default Value | Notes |
---|---|---|---|
Game Thread Priority | 5 | 5 | For the main thread of the client/server. |
Main Worker Priority | 3 | 5 | For the threads created by Main Worker. |
Integrated Server Priority | 5 | 5 | For the client integrated server thread. Does nothing on (actual) servers. |
These are additional optimizations by Smooth Boot.
Option | Smooth Boot Default Value | Vanilla Default Value | Notes |
---|---|---|---|
DataFixer Building | true | false | Datafixer building is very CPU intensive and interferes with how Minecraft normally loads. This optimization forces the datafixers to run on a separate single-threaded executor, which can improve loading times. |