STAR WARS: Galactic Racer on AnyPS5

Screenshots

No screenshot of this game running has been posted in the project's threads yet. Testers who reach a new screen usually post one in the discussion linked on the right.

Where it stands

A single contributor (CnRJay) has been bringing this Unreal Engine 5 game up on Windows over a day. On their combined test branch the game finds all of its system functions, creates the graphics device, runs its UE5 renderer and draws frames. What appears on screen is the boot splash: the Simplygon, Microsoft and Azure PlayFab logos. It is not in the game, and no menu has been reached.

It is also extremely slow, between 0.04 and 0.14 frames per second. The main cause is a very large GPU memory heap that gets copied for every graphics job. Two PRs reduce that cost. One is merged on main and the better one is not yet merged. After a few minutes the game stops on a graphics job that gets garbage size numbers, which is the top blocker at the moment.

Plain main is further behind than the test branch. Many fixes for this game are already merged: missing Sony graphics-library functions, a shader that did not translate, a command-buffer size mistake and a few more. Others are still open or only exist locally. The tester also had to turn off the game's built-in hang watchdog by adding a command-line option, since slow shader builds made the game kill itself on purpose.

Clip distances (parts of the 3D scene that are cut by plane limits) are accepted but ignored, so geometry is not cut correctly even if it did draw.

What is stopping it

• Some compute jobs read garbage group counts, for example 609717307x1x1, causing what looks like a GPU hang after 3-8 minutes. Probably a missing sync between a GPU write and the CPU reading it; the writer is not found yet. Thread
• Performance: a 4 GiB GPU heap (528 MiB actually used) is copied for every dispatch, about 70 GB of reads in 20 seconds. A persistent-mirror fix is open. Thread
• A sparse descriptor heap is copied in full and runs out of GPU memory unless it is cut down to its backed size; the fix is open. Thread
• Clip distances (4 per draw) are not supported on main; the tester ignores them locally. Support is in a draft PR with conflicts. Thread
• One Sony account-service function (sceNpGetSandboxInfo) is only a throwing stub or a local shim, so the real behaviour is unknown. Thread

How it got here

1. 10 Oct 2026Merged: two missing PSSR (Sony's upscaler) functions are declared so the loader stops complaining. Source
2. 10 Oct 2026Merged: a missing audio chat function is declared so the game's bundled party library loads. Source
3. 10 Oct 2026Merged: a graphics fix so the first frame no longer fails on a misread compression metadata pass. Source
4. 10 Oct 2026Merged: shader fixes (an unsupported byte-select instruction and a loop shape that exploded the shader size). Source
5. 10 Oct 2026Merged: fixes for command-buffer building (one-dword NOP, size queries) and branch patching. Source
6. 10 Oct 2026Status post: all imports resolve on the test branch, frames are presented and the boot splash logos show, at 0.04-0.14 fps. Source
7. 10 Oct 2026Opened a PR to mirror the sparse GPU heap instead of copying it for every dispatch, aimed at the 0.1 fps problem. Source

Tested hardware