Persona 5 Royal (PPSA05110)

[thuanquang]

Existing thread

• I searched this category for the title ID and no thread exists for this title on any platform.

Title

Persona 5 Royal (PPSA05110)

Content version tested

01.000.000

AnyPS5 commit

• main fe2b52c7e4 (2026-10-27) with no local changes. It includes fix(libSceGameUpdate): answer no update instead of throwing on request creation #10614, feat(libSceImeBackend): declare 22 sceImeBackend exports #10887 and fix(libSceVideodec2): mark returned pictures as not discarded #10913, and it reaches gameplay.
• Frame rates below come from a look-ahead build, main 73ad397dba with feat(libSceImeBackend): declare 22 sceImeBackend exports #10887, fix(libSceGameUpdate): answer no update instead of throwing on request creation #10614, fix(libSceVideodec2): mark returned pictures as not discarded #10913 and fix(agc): keep at most 4 GiB of the host heap from the host import budget #7844, measured with the GPU to itself.

System tested

Windows 11 26200, Ryzen 5 5600X, Radeon RX 6700 XT (gfx1031), AMD driver 32.0.21045.16010 (Adrenalin 26.10.1), WinLibs GCC 15.2.0, Release.

Furthest point

In game (main, no local changes)

What is shown

• main fe2b52c7e4, with no local changes, gets through:
  • the network prompt,
  • the title screen and New Game,
  • the voice language and disclaimer screens,
  • the opening movie (VP9) and the dialogue after it,
  • first gameplay with free control: "Escape from the casino".
  • Checked on 2026-10-27: no stop in 780 s, with the gameplay HUD on screen 433 s after New Game.
• Picture: frames checked at 960x540 look right.
  • Under the Khronos validation layer, the title run reports one VUID: VUID-VkPipelineColorBlendStateCreateInfo-pAttachments-00605, because independentBlend is not enabled. fix(agc): enable independentBlend for color targets with different blend states #6782 enables it.
  • Output is 3840x2160 (it follows the display).
• Frame rate in the casino: about 15 fps (median 153.9 presents per 10 s over 14 gameplay windows; frame p50 66 ms, p99 88 ms).
  • The GPU is busy 43 ms per present.
  • The draw queue worker is 97 % busy, and profiles show the draw committer thread saturated as well, so draw processing on the CPU limits it.
• Memory: on main, one instance peaks at 21.9 GiB of private memory over that 780 s run, which ends in the casino. On 2026-10-26, a build with the open fix(agc): keep at most 4 GiB of the host heap from the host import budget #7844 peaked at 18.96 GiB on the same route.
  • Without fix(agc): keep at most 4 GiB of the host heap from the host import budget #7844, the host import budget fills at the ATLUS logo, and the refused imports are served by copies. In shared-GPU runs that cost about 3.5 GB more memory and most of the title screen's frame rate.
  • The vkAllocateMemory failure in the first version of this post was Windows running out of commit (RAM plus page file) while other emulator instances shared the PC. It is not a stop in the game, and single runs since then don't hit it.
• Input: works (injected pad input).
• Audio: not checked.
• Beyond the casino escape: not reached yet. Scripted input has not got past it, and no stop is known there.

Current blocker

• Nothing stops main before the casino gameplay.
  • Fixed on main, all merged 2026-10-26: the 22 missing sceImeBackend* exports (feat(libSceImeBackend): declare 22 sceImeBackend exports #10887; this dump never calls them), the update check (fix(libSceGameUpdate): answer no update instead of throwing on request creation #10614), and the divide by zero at the opening movie (fix(libSceVideodec2): mark returned pictures as not discarded #10913).
• Frame rate: the CPU-side draw processing above. The work on it is tracked in the comments below.
• Memory: open fix(agc): keep at most 4 GiB of the host heap from the host import budget #7844 (host import budget, see Memory above) and fix(libkernel): report 12 GiB of direct memory, as a PS5 does #10809.

How it was run

Converted and run with the usage guide's steps. Input was injected with APS5_PAD_INPUT_FILE (scripted pad presses and stick moves) and APS5_NO_PAD_INPUT=1. Frames were checked with APS5_DUMP_FRAMES, and frame rates come from APS5_TRACE_SUMMARY_EVERY=10. Frame rates count only 10 s windows whose frame dumps show the casino HUD, from runs that had the GPU to themselves, without windows where another program used the GPU.

[thuanquang]

Update, 2026-10-26: it now reaches gameplay. I edited the opening post to today's state.

• fix(libSceGameUpdate): answer no update instead of throwing on request creation #10614 is merged, so the update check no longer stops it.
• feat(libSceImeBackend): declare 22 sceImeBackend exports #10887 declares the 22 sceImeBackend exports that main is missing. It is approved. This dump never calls them.
• fix(libSceVideodec2): mark returned pictures as not discarded #10913 fixes the divide by zero at the start of the opening movie: libSceVideodec2 never cleared the picture's discarded flag. It is approved.
• The vkAllocateMemory failure in the first post was Windows running out of commit while other emulator instances shared the PC. It is not a stop in the game, and single runs don't hit it.
• With those three and fix(agc): keep at most 4 GiB of the host heap from the host import budget #7844, it plays the opening movie and reaches "Escape from the casino" with free control.
  • With the GPU to itself, it runs there at about 15 fps (median 153.9 presents per 10 s).
  • Draw processing on the CPU is what limits it.
• One instance peaks at about 19 GiB of private memory there. Two open pull requests lower it, and both were measured on this title:
  • fix(shader): bound the bytes the recompiler's result memos hold #9989 (bounds the recompiler's result memos): about 0.74 GiB less in the casino.
  • perf(libc): keep no per-page physical state for resident guest pages on Windows #11152 (no per-page state for resident guest pages on Windows): about 0.28 GiB less, from boot to the opening.

[thuanquang]

Update, 2026-10-27: #10887 is merged, so main now loads the game. Unmodified main 48c160c58b gets through the title screen and New Game. When the opening movie starts, it stops at the divide by zero that #10913 fixes (0xc0000094 at eboot.exe+0x1347e1a). #10913 is approved, and it is the last stop between main and the casino gameplay. The opening post is updated.

[thuanquang]

Update, 2026-10-27: with #10913 merged, unmodified main fe2b52c7e4 now reaches the first gameplay ("Escape from the casino"), with no local changes. It ran 780 s with no stop, and the gameplay HUD was on screen 433 s after New Game. The opening post is updated. What's left is frame rate and memory, see the post.