mrkeyoor.comAsk

AjncSystem/ShadowOfTheColossus-PCtested 15 Nov 2026commit 6dd820d

ShadowOfTheColossus-PC review

Our sandbox build step failed before launch at a missing directory, while the project reports reaching the credits on v0.60.3.

  • Safety failed
  • Build failed
0 sto install
  • Tests failed
Dev Toolsevaluationupdated 16 Nov 2026
+61stars / 7d

ShadowOfTheColossus-PC is an unofficial Windows port of Bluepoint's PS4 remake, using your own game dump. It runs the game's CPU code directly and translates graphics to Vulkan; it is separate from the PS2 recompilation project.

Verdict

Our sandbox build step failed before launch at a missing directory, while the project reports reaching the credits on v0.60.3. Use ShadowOfTheColossus-PC if you own a supported dump and can troubleshoot the Windows build on your hardware. Skip it if you need an official PC release or a Steam Deck recommendation: the sampled handheld reports describe roughly 12-15 FPS while exploring at 720p.

ShadowOfTheColossus-PC infographic: runs the PS4 remake on your Windows PC, plus our test result and who it suits
Tested 15 November 2026: the source build failed. Before that: source install 0 s.

We ran it

Lab card: what happened when we ran ShadowOfTheColossus-PC
Lab card from our sandbox run
Install✓ · 0s
Build✗ · 0s
Tests✗ · 1sran, no count parsed
Repo2244 files~913,633 lines of source · 66.5 MB · 0 CI workflows · tests dir

Answers from our run

Does ShadowOfTheColossus-PC build from source?

Dependencies installed in 0s, and the build failed. We cloned commit 6dd820d into a clean Debian container with 3 CPUs and no project-specific setup.

Do ShadowOfTheColossus-PC's tests pass?

No. The project's own test command exited unsuccessfully after 1 seconds, and its output did not expose an individual test count.

Who should not use ShadowOfTheColossus-PC?

Players seeking an official Sony PC release: the project is unaffiliated with Sony and Bluepoint and includes no game files.

What are the alternatives to ShadowOfTheColossus-PC?

Oussama268 optimization fork, ShadowOfTheColossus-PC FrameGen, Shadow of the Colossus PS2 recompilation. Our sandbox build step failed before launch at a missing directory, while the project reports reaching the credits on v0.

Setup2/5Our source checkout installed but the build failed.Editor’s note: Not assessed: installation was not reproduced in a supported Windows environment; the sandbox launch path failed.
Docs2/5The 19,750-character README has no usage instructions or examples.Editor’s note: The English README identifies dumps, audio changes and memory profiles; instruction accuracy across all release paths was not assessed.
Community4/5Last push on 2026-11-12, 1,943 stars and maintainers answered or closed 9 of the 10 most-discussed issues we sampled.Editor’s note: Sampled threads show log-driven contributor work and follow-up reports; support capacity and adoption were not measured.
Maturity1/5Latest release v0.75 on 2026-11-12 (pre-1.0), repository created 2026-10-12 and its test suite did not pass in our sandbox.Editor’s note: The project reports completion and continues issuing fixes, but our lab has no gameplay stability result on which to base a rating.

How we score

Who it’s for

Windows players with their own supported PS4 game dump who can track release versions and report hardware-specific problems.
PC gaming enthusiasts willing to compare upstream with a named optimization fork when repeatable streaming stalls interrupt rides.

Who it’s NOT for

Players seeking an official Sony PC release: the project is unaffiliated with Sony and Bluepoint and includes no game files.
Steam Deck owners wanting a comfortable experience from the tested community AppImage: users reported roughly 12-15 FPS exploring at 720p.
Vega owners relying on an older lighting workaround as proof of stability: issue #41 records continuing crashes after colossus fights.

Setup reality

Our November 15, 2026 sandbox run at 6dd820d recorded install success in 0 seconds, build failure (exit 2) in 0 seconds and test failure (exit 66) in 1 second. The build command could not enter /work/home/Games/SotC-PC/SotC-PC before launching Wine. The test tail says 'no test target'.

The Windows release requires your own supported game dump. Version 0.75 supplies a built-in audio mixer, making console sound modules optional. The README recommends putting the game on an SSD. Our 3-CPU, 8 GB sandbox used image label lab-cpp:1; actual toolchain versions require raw logs.

A separate November 16 hands-on attempt stopped because the automation required a Compose file and found none. Its summary's deployment wording conflicts with that failed step. Teardown succeeded. Neither record contains a completed game launch.

The project reports credits; our launch stopped before gameplay

ShadowOfTheColossus-PC runs the PS4 remake's game code on a Windows PC, and the project reports reaching the ending and credits on v0.60.3. Use it if you already have your own supported game dump and enjoy working through version-specific PC gaming problems. Skip it if you want a Sony-supported PC release or a Steam Deck recommendation backed by good reported frame rates. Our sandbox stopped at a missing launch directory, before gameplay.

The upstream README describes a loader: the PS4 game's existing x86-64 instructions execute on your CPU, system functions get replacements, and graphics work goes through Vulkan. That distinction matters when searching for downloads. This is AjncSystem's port of Bluepoint's PS4 remake, not the similarly named project that recompiles the PS2 original into C++.

Version 0.75 removes a significant setup obstacle: its built-in mixer makes console sound modules optional. The game dump is still your responsibility. The supported identifiers in the README are CUSA08809 EU 1.01 and CUSA08034 US 1.00. Start by matching those identifiers and the release version, before copying settings from a video or an old fork's instructions.

What people use ShadowOfTheColossus-PC for

The useful reports cover finishing the game, improving older CPUs and trying community Linux builds, with very different outcomes.

Who used it Build and task What they reported
Project team v0.60.3, last colossus through the ending Reached the credits with high-quality textures, PS4 Pro mode and 60 FPS enabled. This is the project's completion report.
Oussama268's release author v0.60.6-opt11 on an RTX 3060 Laptop A 30-minute session averaged 57.5 FPS with a 40.1 FPS 1% low. A separate rides-and-reloads session reported no crash or rainbow textures. Release test
GitHub user LHBSvaz v0.60.6-plus26, i5-3470 and RTX 2060 Defeated colossi 3, 4 and 5 at around 25-30 FPS after a polling-sleep change. An earlier session on plus26 crashed during area loading after about 43 minutes. Issue #267
GitHub user Ivanprestef2 Vega 7 integrated graphics, experimental lighting changes Defeated a colossus after the lighting improved, then reported crashes, especially after victories. Issue #41
GitHub users susieltarune and Emanu1995 Community AppImage on Steam Deck Reported low frame rates at 720p, with the second user describing similar performance. Issue #23

LHBSvaz's account is the most useful antidote to a simple minimum-spec list. The earlier thread-priority build still performed poorly on that user's actual machine. A later change let waiting job threads sleep instead of continually polling for work, and the user then reported playable fights. That sequence gives an older-CPU owner something specific to investigate. It also shows why a result from a modern CPU restricted to fewer threads should not replace a report from the older processor itself.

The Vega account makes a different point: reaching and defeating a colossus is useful evidence, but the crash after the fight matters to someone trying to keep playing. Treat a successful lighting workaround and a stable play session as separate questions. Both outcomes appear in the same user's thread.

What happened when we ran it

Our November 16 hands-on attempt stopped at the safety step because the repository had no Compose file. The step record says the required compose.yaml, compose.yml, docker-compose.yaml or docker-compose.yml was absent. That is a mismatch with this automation path, not evidence that a desktop game ought to ship as a container service.

The run-summary record describes deployment with the repository's own Docker Compose, but the failed step contradicts that description. We count this attempt as blocked, rather than a working deployment. Teardown reported success. No gameplay question or screenshot accompanies this attempt.

Our separate sandbox run used commit 6dd820d on November 15, 2026, with 3 CPUs, 8 GB RAM and image label lab-cpp:1.

Sandbox step Recorded result What the log supports
Install Succeeded in 0 seconds The harness recorded a successful step.
Build Failed, exit 2, in 0 seconds The game target could not change into /work/home/Games/SotC-PC/SotC-PC.
Tests Failed, exit 66, in 1 second The tail says "no test target".

The build command tried to enter that directory and run wine SotC.exe --run. The shell stopped at cd: it could not enter the directory. Wine never got a chance to launch the executable in that command. Calling this a compiler failure would misdescribe the evidence, and blaming a graphics driver would go beyond it. The recorded build step failed before either question arose.

Likewise, "no test target" is a harness outcome, not a count of failed game tests. The scan found a tests directory, zero CI workflow files and no Dockerfile. It recorded 2,244 files, approximately 913,633 lines of source and a 66.5 MB checkout. Those observations help describe the checkout, but they do not turn the missing test target into a test suite result. The image label supplies no actual compiler, runtime or operating-system version.

Where people get stuck: streaming, CPU scheduling and AMD graphics

Repeated area-entry stalls need a different investigation from shaders encountered for the first time. In issue #21, Lucas-blip-png asks the reporter to walk the same route twice: a stall that repeats on the second pass points to area streaming rather than first-use shader compilation. Tasosgemah then agrees that their freezes are streaming-related. That is a practical diagnostic you can use before spending an evening chasing a shader-cache setting.

Symptom What the discussion found Reported outcome or next step
Immediate close on an older CPU Lucas-blip-png identified missing BMI1, LZCNT and MOVBE instructions in #88, correcting the early AVX2 explanation. Emulation coverage expanded in 0.59-plus5, followed by a real-hardware confirmation. AVX remains required. Issue #88
Four-thread CPU performs badly LHBSvaz still saw poor results after the priority change. The polling-sleep change in plus26 produced a reported improvement and completed fights. AjncSystem later said the work was incorporated. Issue #267
Vega lighting improves but the game crashes Ivanprestef2 reported improved lighting with the experimental wave32 path. Disabling it brought visual problems back and did not eliminate the reported crashes. Issue #41
Linux GCN workaround crashes during the intro albaruzz corrected an earlier interpretation: the requested 32-wide compute mode was invalid on that RX 580 driver path. Lucas-blip-png said plus5 disables the default path where the driver cannot require that compute subgroup size. Issue #201
Video-memory exhaustion or device loss #123 contains both memory-capacity discussion and RX 570 GPU-hang logs. Lower settings and allocation changes helped some paths. The RX 570 reporter still described crashes and returning visual bugs. Issue #123

Do not turn that last row into "all 4 GB cards are broken" or "the memory leak is fixed." In the same thread, dougbug589 reports a successful short Linux run on a GTX 1660 Ti 6 GB and explicitly separates it from the AMD case. Different GPUs, drivers and workarounds are doing different things. A useful bug report names the exact build, GPU, settings and point in the game.

Fewer hitches can cost more RAM without raising FPS

Oussama268's v0.75-opt15 ride benchmark reports the same FPS, fewer long hitches and 200-500 MB more RAM use. On the author's RTX 3060 Laptop test, frames over 100 ms fell from about 5.5 to 2 per run, while the worst frame fell from about 140 to 110 ms. These are the fork author's measurements, not ours. Optimization fork.

That trade is more informative than another headline promising higher FPS. A ride can feel better when interruptions shrink even if its average frame rate stays put. It can also still interrupt you: the reported worst frame remains above 100 ms. Compare the same route and settings when choosing a fork, and look at long frames alongside the FPS counter.

The upstream README recommends an SSD because new areas stream textures and models from the game folder. It also documents an automatic light profile for cards with less than 8 GB of video memory from v0.72.1. Leave that profile in the decision: a screenshot made with a larger card's texture settings tells a small-card owner little about the experience they should expect.

Game ownership and release identity come before settings

The repository includes no game or Sony files, so downloading the port alone does not give you Shadow of the Colossus. The current README requires your own game dump and says the project is unaffiliated with Sony Interactive Entertainment or Bluepoint Games. That boundary is central to setup, including when a download page calls the package a native PC port.

Use the upstream repository and the named fork's own releases to identify what you are evaluating. This is particularly relevant for old instructions about copying console audio modules: they describe a requirement that v0.75 changed. The latest upstream release also warns that the v0.73 and v0.74 launchers checked the wrong update location, requiring one manual download of a corrected release. A launcher that offers no update can therefore be misleading. v0.75 release notes.

GitHub's public metadata on November 16 showed 1,943 stars, a November 12 last push and v0.75 published November 12. Stars measure interest, not completed playthroughs. The issue extracts supply more useful evidence of community work: users posted logs, contributors issued targeted builds, and users returned with both successes and crashes. That activity supports trying the project with a specific problem in mind, rather than assuming every recent fix covers every GPU.

Windows is upstream; native Linux builds are community work

The sampled maintainer reply in issue #23 says there is no official Linux build. Community AppImages and Windows builds run through Proton are separate choices. FlaipyTheHost reports good results on Bazzite with an AMD GPU, while Toilet-Lord and ACheshirov report successful Proton use in issue #402. Those accounts establish working routes for those users, without a controlled comparison between them.

The English README gives Windows users an entry point, while the issue discussions mix English and Portuguese. Older CPU support in #88 is discussed largely in Portuguese. Keep exact identifiers such as BMI1 and MOVBE when searching those threads, because an English-only search for "AVX2 fix" can lead you back to the explanation the contributor corrected. The v0.75 notes also say the launcher's What's New screen shows English notes for languages other than Portuguese.

For Steam Deck, the reports are much less encouraging than the existence of an AppImage suggests. Native packaging makes a program easier to launch; it cannot by itself establish a good handheld experience. The reported frame rates and the earlier launcher packaging mistake deserve their own answers below.

ShadowOfTheColossus-PC vs Oussama268, FrameGen and the PS2 recompilation

Choose between these projects by game version and the specific change you need, because the forks do not represent interchangeable downloads.

Project Reason to choose it Reason to be careful
AjncSystem upstream The PS4 remake loader with v0.75 built-in audio and a history of merging community work. Match each reported fix to your build and hardware.
Oussama268 optimization fork v0.75-opt15 specifically targets area-entry hitches. Its author reports extra RAM use at unchanged FPS, and earlier optimization work has moved upstream. Fork
nakaionyt FrameGen fork Its older v0.57-based package focuses on presenting 60 frames from a 30 FPS game cap. Its README explicitly lacks a completed full-game check for that exact build. Newer upstream also includes frame generation. FrameGen
volstants recompilation You want the PS2 PAL original rather than the PS4 remake. It statically recompiles a specific PS2 disc image into C++ and has its own local build requirements. PS2 project

My first choice for a new Windows evaluation would be upstream, followed by the optimization fork if repeatable riding stalls are the problem. That recommendation follows the release identities and the fork's stated purpose. It is not a comparative performance result from our lab. Keep saves backed up and retain the build identifier when comparing sessions so a better result can be traced to the version you actually ran.

What the other reviews get wrong

FramesGames and HotRepo's mandatory-console-audio advice predates v0.75's default built-in mixer. Both pages describe needing console sound files, while the later upstream README makes them optional. Apply that correction to v0.75, rather than assuming an older fork acquired the new mixer too. FramesGames, HotRepo.

HotRepo also limits its description to EU CUSA08809 and excludes Maximum textures from PS4 Pro mode. The later upstream history documents US CUSA08034 support and the addition of Maximum textures in Pro mode, initially experimental. Its blanket GCN boundary also needs the later v0.74.2 colour-fix notes alongside the separate crash reports. A colour correction and a crash correction are different claims.

GameGPU_com's X post calls the PS4 port static recompilation. The project description instead calls it a loader executing the existing x86-64 game code. Static PS2-to-C++ recompilation describes volstants' separate project. The same X coverage calls rendering native 4K, whereas the supplied FramesGames guide describes 1080p rendering scaled to the selected output. These records support distinguishing output resolution from internal rendering, rather than repeating the native-4K claim. X records, PS2 project, setup guide.

The Quiver fork listing also has an internal version disagreement: its release label says v0.74.1-opt14 while its embedded README names v0.75-opt15. Follow the named repository's release information when that happens. A directory listing's "latest" badge is a poor substitute for identifying the executable in your own session. Fork listing.

Questions people ask

The ShadowOfTheColossus-PC GitHub repository is AjncSystem/ShadowOfTheColossus-PC, the project identified in the supplied upstream listing.

Where is the ShadowOfTheColossus-PC GitHub repository?

The upstream code and releases are at AjncSystem/ShadowOfTheColossus-PC. Oussama268's similarly named repository is an optimization fork, and volstants' project targets the PS2 original. Check the owner as well as the repository name. PS2 identity.

Is this PC port real or fake?

AjncSystem publishes a PS4-remake loader with source code, releases and community gameplay reports. The project's README reports reaching the credits, and LHBSvaz reports completing specific colossus fights. That is stronger evidence than the X reply asking whether the port is real. Download identity still matters. user report.

Is Shadow of the Colossus officially available on PC?

This project is an unofficial PC port, unaffiliated with Sony or Bluepoint. The supplied FramesGames report says Sony never brought the PS4 remake to PC. Do not read the words "native Windows port" as an announcement from the game's publisher. platform report.

Can you finish the entire game with this port?

The project reports completion through the ending and credits on v0.60.3 with high-quality textures, PS4 Pro mode and 60 FPS enabled. That is a project-reported result tied to that build and configuration. The older FrameGen fork separately says a full playthrough on its exact build has not been checked. FrameGen boundary.

Why does the game have no sound?

Older builds needed console sound modules; v0.75 adds a built-in mixer enabled by default. If you are following older instructions, identify the release first. The new Sound page also offers device and channel choices, and console modules remain an optional engine choice.

Can the port support older CPUs that lack newer instruction sets?

The community added emulation for BMI1, LZCNT and MOVBE, with real-hardware confirmation recorded in issue #88 after expanded coverage in 0.59-plus5. The contributor says AVX remains mandatory. The thread's early suggestion that AVX2 explained the failure was corrected later in the discussion. Older CPU investigation.

Can thread scheduling changes make the game playable on a four-thread CPU?

LHBSvaz reports around 25-30 FPS during colossus fights on an i5-3470 after the plus26 polling-sleep change. The earlier priority-only build still performed badly for that user. This is a specific success report, not a promise that every four-thread CPU will reach that frame rate. Scheduling investigation.

Will there be an official native Linux or AppImage release?

The sampled upstream reply says there is no official Linux build and points users toward community-maintained Linux work. Issue #402 links a community Linux release and contains Proton success reports. Neither thread supplies a commitment to an official AppImage release. Support boundary, community routes.

How do I open the launcher in the Linux AppImage?

In issue #23, FlaipyTheHost directs affected users to the v0.57.1 AppImage after acknowledging a packaging mistake. Susieltarune then confirms the launcher works. That answer belongs to that community package: use its corrected release rather than assuming every AppImage opens the Windows launcher's interface. Launcher exchange.

How well does it run on Steam Deck, and what FPS can it reach?

Susieltarune reports roughly 12-15 FPS exploring at 720p, occasionally 20 FPS, and about 10 FPS looking at the Shrine. Emanu1995 reports similar performance. These dated community results give us no reason to recommend that tested setup for comfortable handheld play. Steam Deck reports.

Is there an option or INI toggle for shader preload at startup?

Issue #21 describes preparation of cached shaders at startup, but its final commenter still asks where the preload option is and reports continuing stutters. The earlier contributor reply distinguishes cached preparation from full first-run precompilation. That exchange supplies no confirmed toggle to recommend. Shader discussion.

Where can I find SCSKiller, and does it help with these freezes?

Issue #21 mentions SCSKiller, but the reporter says it did not help and another user asks for a link. The extract supplies no confirmed download source or successful fix. Use the thread's repeat-the-route diagnostic to distinguish streaming stalls before treating an unrelated utility as a remedy. SCSKiller exchange.

What we have not tested yet

Our lab has no completed gameplay run of this port: the November 15 command stopped at a missing directory, and the November 16 automation stopped at its Compose requirement. We have not tested audio quality, controllers, save recovery, full-game stability, native Linux, Proton, Steam Deck or current AMD fixes. The actual sandbox toolchain versions require the raw logs. Neither attempt supports a frame-rate claim, a security assessment or a passing-test count.

The independent evidence gap is a current-build playthrough and a controlled comparison across hardware. Project completion reports, developer benchmarks and individual user sessions are useful, but their builds and settings differ. The sampled closed issues sometimes end with a request to retest v0.75 rather than a fresh result from the affected machine. In particular, the later GCN colour-fix notes do not resolve the older Vega user's crash history by themselves.

The next useful check is a supported Windows release with a matching dump, followed by the same ride repeated and a colossus fight carried through its save transition. Record the build, hardware, texture profile and long frames together. A smooth opening cinematic would not settle the problem these users encountered after a victory.

Alternatives

ProjectWhat it isPick it when
Oussama268 optimization forkA PS4-remake fork whose v0.75-opt15 targets area-entry hitches.Try it when repeatable riding stalls are your problem and the reported extra RAM use is acceptable.
ShadowOfTheColossus-PC FrameGenAn older v0.57-based fork focused on frame generation from a 30 FPS game cap.Choose it for that specific pacing experiment, accepting its stated lack of a full playthrough check on the exact build.
Shadow of the Colossus PS2 recompilationA separate project statically recompiling the PS2 PAL original into C++.Choose it when you want the PS2 original and have its required disc image, rather than the PS4 remake.

What people are saying

  1. [velocity-scout] AjncSystem/ShadowOfTheColossus-PC

Sources

  1. Upstream README and source
  2. Version 0.75 release
  3. Optimization fork
  4. FrameGen fork
  5. PS2 recompilation project

More dev tools reviews