Skip to content
SPECIFICTION
Get listed

FerriteCore

github.com

Free plan Self-hosted
Visit github.com
Spec score 8.3 #3 of 24 in Minecraft Performance Mods

What it is

FerriteCore is a free, MIT-licensed coremod for Minecraft: Java Edition that targets RAM use. It reduces memory associated with block states, models, predicates, strings, voxel shapes and baked quads through a set of documented optimizations. Examples include replacing block-state neighbor tables with FastMaps, caching multipart model predicates, deduplicating model instances and reducing baked-quad memory. The project reports that individual techniques saved hundreds of megabytes in cited tests; one pack example showed post-garbage-collection memory moving from roughly 3.1 GB to 1.1–1.2 GB on the title screen. Several optimizations have no CPU impact during play, while some add load-time CPU cost. ThreadingDetector optimization is disabled by default because it may very rarely cause race conditions. FerriteCore supports Fabric, Forge, NeoForge and Quilt and is listed for Minecraft Java Edition versions 1.16.5 through 26.3, depending on loader builds. It is marked for both client and server use, and the maintainer recommends installing it on both because some optimizations apply server-side. Versions before 1.16 are not supported for backporting.

Spec sheet

Purpose
FerriteCore is a coremod that reduces Minecraft's memory usage through multiple optimizations.Read 1 Oct 2026 · github.com
Support
The project links to GitHub for reporting issues and viewing source code.Read 1 Oct 2026 · modrinth.com
License
FerriteCore is distributed under the MIT License.Read 1 Oct 2026 · github.com
Block-state savings
Replacing vanilla block-state neighbor tables with FastMaps is documented as saving around 600 MB in the cited test.Read 1 Oct 2026 · github.com
Blockstate neighbors
Replacing per-state neighbor tables with FastMaps saves around 600 MB in the documented test.Read 1 Oct 2026 · github.com
Compatibility
The listed Minecraft Java Edition compatibility includes versions from 1.16.5 through 26.3, depending on loader builds.Read 1 Oct 2026 · modrinth.com
Fabric benchmark
In All Of Fabric 3 version 2.5.9, heap usage fell from 1,792 MB to 984 MB after generating a world and waiting two minutes.Read 1 Oct 2026 · curseforge.com
Installation guidance
The maintainer recommends installing FerriteCore on both client and server because some high-impact optimizations apply to the server.Read 1 Oct 2026 · curseforge.com
Known incompatibility
The Fabric page says FerriteCore builds for Minecraft 1.18 and newer are incompatible with Hydrogen because Hydrogen is archived and receives no releases for those versions.Read 1 Oct 2026 · curseforge.com
Memory reduction
In the 1.16.4 Direwolf20 pack example, memory after garbage collection fell from roughly 3.1 GB to 1.1–1.2 GB on the title screen.Read 1 Oct 2026 · curseforge.com
Mod loaders
Supported platforms listed are Fabric, Forge, NeoForge and Quilt.Read 1 Oct 2026 · modrinth.com
Model instances
Deduplicating multipart model instances reduces about 200,000 instances to about 1,500 and saves close to 200 MB.Read 1 Oct 2026 · github.com
Model predicates
Caching multipart model predicates saves 300–400 MB in the documented test.Read 1 Oct 2026 · github.com
Model savings
Caching multipart model predicates is documented as saving 300–400 MB in the cited test.Read 1 Oct 2026 · github.com
Optimization focus
The mod optimizes memory used by block states, models, predicates, strings, voxel shapes and baked quads.Read 1 Oct 2026 · github.com
Optimization techniques
The project documents optimizations including Optional cleanup, block-state neighbor storage, property storage, predicate caching, model deduplication, shape deduplication, quad deduplication, and threading-detector changes.Read 1 Oct 2026 · github.com
Property storage
Replacing ImmutableMap blockstate property storage saves around 170 MB in the documented test.Read 1 Oct 2026 · github.com
Quad deduplication
Quad deduplication reduces baked-quad memory by close to 150 MB in the documented test.Read 1 Oct 2026 · github.com
Runtime behavior
Several documented optimizations have zero CPU impact while playing, with some load-time CPU cost.Read 1 Oct 2026 · github.com
Runtime environments
The project is listed as supported on both client-side and server-side environments.Read 1 Oct 2026 · modrinth.com
Supported environments
The project is marked as supporting both client-side and server-side environments.Read 1 Oct 2026 · modrinth.com
Supported game
Modrinth lists compatibility with Minecraft: Java Edition versions from 1.16.5 through 26.3.Read 1 Oct 2026 · modrinth.com
Supported loaders
The project is listed for Fabric, Forge, NeoForge, and Quilt.Read 1 Oct 2026 · modrinth.com
ThreadingDetector
The ThreadingDetector optimization is disabled by default because it may cause very rare race conditions.Read 1 Oct 2026 · github.com
Version limit
The maintainer says FerriteCore will not be backported to Minecraft versions before 1.16.Read 1 Oct 2026 · curseforge.com

Plans and prices

FerriteCore open-source mod Free MIT License · Minecraft Java Edition mod Read 1 Oct 2026 · modrinth.com

Every FerriteCore plan and price →

Strong points and limits

Strong points

  • Targets memory used by models and block states.
  • Supports Fabric, Forge, NeoForge and Quilt.
  • Listed for both client and server.
  • Free under the MIT license.

Limits

  • Will not be backported to versions before 1.16.
  • Some optimizations add load-time CPU cost.
  • ThreadingDetector changes may rarely cause race conditions.

Who it suits

FerriteCore suits Minecraft: Java Edition users looking to reduce memory use and able to select a compatible loader build. The maintainer recommends installing it on both client and server.

Verdict

FerriteCore documents its memory-focused techniques and supports several loaders across a broad listed version range. Check version and loader compatibility, and note the recommendation to install it on both client and server.

Where Specifiction places it

Alternatives to FerriteCoreAll alternatives →