Table of Contents

Stride 4.4 Release Notes

Stride 4.4 is one of the largest engine updates in years, with roughly 2,400 commits since 4.3. The theme of this release is modernization and reach: much stronger Vulkan and Direct3D 12 backends, full platform coverage across Windows, Linux, macOS, Android and iOS (all continuously tested on CI), a from-scratch shader compiler, and a real command-line workflow that no longer needs Game Studio.

๐Ÿ“ท [Hero image, e.g. a Stride scene rendered through the new shader pipeline]

โœจ Highlights

๐Ÿ“ฑ Platforms & cross-platform

Stride 4.4 dramatically widens both where you can build games and where they run.

Build anywhere: the asset compiler and the full engine build now run on Linux and macOS, not just Windows.

Run everywhere: each target was brought back into shape (re-implementing whatever had bit-rotted) and locked down with per-platform CI, so it can't silently break again:

  • Windows โ€” Direct3D 11/12 and Vulkan.
  • Linux โ€” full runtime + tooling.
  • macOS โ€” net10.0-macos runtime, Vulkan via MoltenVK, AVFoundation video and audio.
  • Android โ€” building and running reliably again after a round of fixes.
  • iOS โ€” builds restored, with platform defines and orientation handled correctly.

A Stride sample running on a physical iPhone

โŒจ๏ธ A real command-line workflow: the stride CLI + dotnet new templates

You can now create, build and run Stride games entirely from the command line, with no Game Studio install required.

The new stride dotnet tool bundles a full project lifecycle into a single, scriptable, CI-friendly command that effectively replaces the old Launcher:

Command Purpose
stride new scaffold a project from installed templates (stride new list to browse)
stride build / stride asset build the project and compile assets
stride upgrade upgrade a project to a newer Stride version
stride studio launch Game Studio for the resolved version
stride sdk install / uninstall / update manage installed Stride versions
stride self update update the CLI itself
stride version report resolved versions 
dotnet tool install -g stride.cli --version 1.0.0 # install the Stride CLI
stride sdk install                                # install the latest Stride
stride new topdownrpg && cd TopDownRPG            # create a project from a template
stride studio                                     # open it in Game Studio

dotnet new templates are also available as a lightweight fallback if you'd rather use the standard .NET tooling directly:

dotnet new install Stride.Templates
dotnet new stride-game -o MyGame

๐ŸŽฎ Vulkan & Direct3D 12

Both backends got a big overhaul and stability pass this cycle and are in much better shape. They're now solid enough that we expect to make a modern backend the editor default before long, and Direct3D 11 is a candidate for removal in the next major release. GPU crashes are also far easier to track down: Stride can now pinpoint the exact rendering step that caused a device hang.

If you write custom low-level rendering code, note that D3D12 and Vulkan now use an explicit barrier/layout model (and D3D12 requires Enhanced Barriers โ€” the legacy path was removed).

You can also pick the graphics API right from the UI now. Game Studio itself can render with a chosen backend via Settings โ†’ Environment โ†’ Graphics API (Game Studio only) โ€” this takes effect after a restart:

Choosing the Game Studio graphics API in Settings โ†’ Environment

And each Windows game project can select its own Graphics API from the package build settings in the property grid:

Selecting a Windows project's graphics API from the property grid

๐ŸŽจ A brand-new SDSL shader compiler

The biggest internal change in 4.4 is a complete, from-scratch rewrite of the SDSL shader compiler, now built around a modern SPIR-V-centric pipeline.

Instead of parsing and stitching shaders together as text, Stride now works in SPIR-V bytecode end to end:

  • Each .sdsl shader is parsed once and compiled into its own SPIR-S module (SPIR-Stride, Stride's extended SPIR-V dialect).
  • Effects (.sdfx) then mix and compose those modules directly as bytecode, lowering the result to standard SPIR-V for the GPU backend.
  • Crucially, text parsing happens only at that first step: recombining a new shader variation from already-compiled SPIR-S needs no re-parsing.

What this means for you:

  • Much faster shader handling. Generating the many shader permutations a real game needs no longer touches a text parser; variations are recombined straight from cached bytecode.
  • Built on mature, standard tooling. Vulkan consumes the SPIR-V natively, while Direct3D 11/12 and Metal reuse SPIRV-Cross (to HLSL/MSL), battle-tested components instead of a bespoke translator.
  • Far better support for advanced features. Direct3D 12 and Vulkan now handle things like tessellation and compute shaders much more reliably.
  • A future-proof foundation. With a real SPIR-V pipeline in place, adding modern GPU features such as ray tracing, mesh shaders/meshlets and wave intrinsics becomes much easier going forward.

โš ๏ธ Heads-up: Because the entire shader compiler was replaced, custom .sdsl shaders may need minor adjustments to compile cleanly. If you maintain shaders, give them a pass on 4.4. See Upgrade Notes below.

The new SDSL shader pipeline: many .sdsl shaders are parsed once into per-shader SPIR-S bytecode, .sdfx effects mix and compose them into standard SPIR-V, which feeds Vulkan natively and Direct3D and Metal via SPIRV-Cross

Huge thanks to Youness Kafia, whose early prototyping and experimentation laid the foundation for the new SDSL pipeline.

โšก NativeAOT- and trim-clean engine

The engine is now NativeAOT- and trimming-friendly: a full AOT publish completes with 0 errors. This unlocks smaller, faster-starting, self-contained game builds.

๐Ÿ›  Engine & Graphics

  • Video subsystem rewrite: a fresh Windows Media Foundation backend, plus a new AVFoundation backend on macOS.
  • Direct3D 11 stability & correctness fixes.
  • Fonts: upgraded to FreeType 2.13.
  • Model importing: upgraded to Assimp 6.

๐Ÿงฐ Build, Tooling & Project System

  • Much faster asset builds. Warm asset compiles drop by around 40% for a typical game, and up to 10ร— for Stride's own tests, thanks to a new asset-build cache.
  • .slnx is the new default solution format for projects created by Stride. Existing .sln solutions still open and save normally.

๐Ÿงช Quality & CI

Mostly under the hood โ€” but it directly changes how confidently you can contribute back to the engine.

Stride 4.4 has by far the most robust CI the project has ever had. Where earlier versions only exercised a slice of platforms, every change now runs the entire test matrix in one pipeline โ€” engine builds, game and Game Studio tests, and end-to-end sample/packaging builds, across all platforms and graphics APIs, including GPU image-comparison tests (Windows D3D11/12, Linux Vulkan, macOS, Android, iOS):

The CI pipeline: a single run building and testing every platform and graphics API (Windows D3D11/D3D12/Vulkan, Linux, macOS, Android, iOS), all green

Breakage on any platform or backend is now caught automatically before it reaches a release. For contributors that's the real win: you can open a pull request and trust CI to prove it works everywhere, instead of testing each platform by hand โ€” which makes contributing a feature back to the engine far less daunting.

A gold-image generation workflow runs directly on CI, so you no longer have to regenerate reference images by hand on every platform (or wait for a long test run to fail before noticing) โ€” golds are produced and promoted straight from the CI workflow.

The new CompareGold tool makes reviewing these tests painless โ€” visually diff failures against their gold images, promote the ones you accept, and even pull results directly from any CI run or fork (see GPU-TESTING.md).

CompareGold reviewing image differences and promoting gold images

๐Ÿ› Notable Fixes

  • Game Studio no longer hangs with the main window left disabled after a project upgrade.
  • Fixed EffectEvaluator for nested composition arrays.
  • Fixed Plane.Negate and several null-reference and internal calculation bugs.
  • Fixed D3D12 swapchain creation and a range of D3D11 stability issues.
  • Numerous shader-compiler correctness fixes.

โฌ†๏ธ Upgrade Notes

  • .NET 10 SDK is required to build the engine and games (same as 4.3).
  • The standalone Launcher workflow is superseded by the new stride CLI.

Breaking changes

  • Custom shaders: the SDSL compiler was rewritten, so review and recompile custom .sdsl/.sdfx shaders; some non-standard constructs may need small fixes. If you hit a shader that no longer compiles or behaves differently, please open an issue on GitHub so we can fix it.
  • Direct3D 12 now requires Enhanced Barriers; the legacy barrier path has been removed.
  • Stride.Core.Serialization no longer depends on MicroThreading (#3169): code that relied on this transitive reference must now reference the MicroThreading assembly directly.
  • Bepu CharacterController was reworked on top of Bepu's own controller (#3195): existing character setups may behave differently and need adjustment.
  • Bepu convex hull behavior changed: existing convex hull colliders may need review.

๐Ÿ™ Contributors

Thanks to everyone who contributed to Stride 4.4:

  • Youness Kafia (ykafia)
  • Nicolas Musset (Kryptos-FR / Color-Rise)
  • Mario Guerra
  • Johan Gustafsson
  • Eideren
  • Vaclav Elias
  • Virgile Bello (xen2)
  • Jakub ลawreszuk
  • Feralnex (Tomasz Czech)
  • Nicogo (Nicolas Gomez)
  • Doprez
  • ds5678
  • w0wca7a
  • Basewq
  • D4rkDuck
  • Elias Holzer
  • Peter Laske
  • Will Bentz
  • Henrik Gedionsen
  • Kevin Norris
  • Luca Domenichini
  • MiyaGrace
  • MsEpsilon
  • Rafael Stahl
  • Redwarx008
  • Steve Berdy

โ€ฆand everyone who reported issues, tested builds and helped on the community channels. ๐Ÿ’™

Edit this page