Quick Answer: The GTA 6 PC port on GitHub, called gta6recomp, is an experimental community framework, not an official Rockstar release. It targets DirectX 12 shader stutter, asset streaming, and frame pacing. Its README reports 18–22% FPS gains in one internal benchmark. None of those numbers are independently verified.
GTA VI on PC is one of the most-searched gaming topics right now, and a repo promising a "PC port" is going to attract attention. Shader stutter is the problem most PC players dread: you walk into a new street and the frame time spikes for half a second. The gta6recomp project says it can fix exactly that. Before you download anything, you need to know what it does, what it doesn't prove, and where the risks are.
What the gta6recomp Project Claims to Do
The GitHub README describes version 1.4.0 as a "runtime shader recompilation and performance optimization framework." It lists three areas of work:
- Rendering: runtime shader cache generation, dynamic Pipeline State Object (PSO) management, and multi-threaded shader preloading.
- CPU: dynamic thread management, hybrid-core scheduling for Intel and AMD processors, and background task balancing.
- Memory and streaming: texture streaming changes, asset loading prioritization, and dynamic decompression scheduling.
The PSO work matters most. In DirectX 12, a PSO bundles a compiled shader with its render state. If the game compiles a PSO the first time it's needed, the frame stalls while the driver does the work. Preloading those PSOs in background threads is a real, established technique, and it's the core claim here.
Notice what the README doesn't say. It never explains how the framework interacts with the game binary, and that's the first question any experienced modder would ask. Until that's documented, treat the feature list as a statement of intent.
Why Shader Stutter Happens, and the Advice Most Guides Get Wrong
Here's the mechanism. When the GPU encounters a material or effect it hasn't compiled yet, the engine requests a PSO. The driver compiles it synchronously, and the render thread waits. A single compile may take a few milliseconds, but a dense city can request dozens in a second.
The common advice is "add more CPU threads" or "turn on a faster preset." That misses the point. Compilation stutter is about when work happens, not how much total work there is. Adding threads helps only if the compile is moved off the render path.
Here's the counter-intuitive part. A shader cache build on first launch makes the first session look worse, not better. The framework spends time compiling everything up front, so anyone benchmarking the first hour will get misleading numbers. Wait until the cache is warm before you judge performance.
This next part matters more than it looks: the cache is hardware- and driver-specific. A cache built on one GPU driver version may be invalid after an update, so expect rebuilds after driver changes.
Reading the Benchmark Table Honestly
The README includes this table, measured in what the author calls an "internal benchmark" of a dense urban environment with maximum traffic and pedestrian density:
| Configuration | Native FPS | Optimized FPS | Gain |
|---|---|---|---|
| RTX 3060 | 58 | 71 | +22.4% |
| RTX 4070 | 87 | 104 | +19.5% |
| RTX 4080 | 116 | 138 | +19.0% |
| RX 7900 XTX | 122 | 145 | +18.9% |
The gains are real arithmetic on the author's numbers. The problem is what's missing. The README doesn't disclose the scene path, the resolution, the upscaler setting, the driver versions, or how many runs were averaged. It reports average FPS only, with no 1% lows and no frame-time graphs. Average FPS can rise while stutter stays the same, which is exactly the problem this project says it solves.
That said, the pattern is plausible. Gains are largest on the lower-end card, which fits a bottleneck relief story. Still, a roughly 20% uplift from a community tool with no published methodology should be read as a claim to test, not a fact to rely on.
Hardware Requirements and What They Imply
The README lists two tiers. The minimum is a Ryzen 5 5600 or Intel i5-12400, 16 GB of RAM, an RTX 3060 or RX 6700 XT, an SSD, and Windows 10 22H2. The recommended tier is a Ryzen 7 7800X3D or Intel i7-14700K, 32 GB of RAM, an RTX 4080 or RX 7900 XTX, an NVMe SSD, and Windows 11 24H2.
Read those tables as a hint about where the bottleneck is. The framework leans on CPU threading and streaming, so the recommended CPU and NVMe drive are not decoration. If you have 16 GB of RAM and a SATA SSD, the streaming improvements are the ones most likely to disappoint you. Upgrading RAM before testing the mod is the cheaper experiment.
Upscaling and Frame Generation: Which Path to Pick
The project lists support for NVIDIA DLSS, AMD FSR, and Intel XeSS. Each has different hardware rules, and the choice affects your results more than the mod does.
- DLSS runs on RTX GPUs with dedicated tensor cores and generally gives the best image quality for its cost.
- FSR works across most modern GPUs, including Nvidia, which makes it the most flexible option.
- XeSS uses XMX matrix units on Intel Arc cards and falls back to a DP4a path on other hardware, so quality and speed vary by GPU.
Here's the trade-off that catches people out. Frame generation raises the displayed FPS but doesn't reduce input latency, so a 140 FPS number can feel like 70 in a fast fight. If you're testing the mod, turn off frame generation first. Otherwise you'll confuse a rendering improvement with an interpolation effect.
Installation: The Part That Needs the Most Caution
The README's install steps are short. Extract the archive into the game directory, run Setup.exe, launch the game, and let the shader cache build. Logs go in the logs folder, and the expected layout includes cache/, profiles/, shaders/, GTA6.exe, and config.ini.
Before you do any of that, take these precautions:
- Keep a clean copy of your game install. Back up the original folder or use a second install so you can roll back.
- Read the code before running any binary. Setup.exe comes from an unsigned GitHub archive, so review the source or build it yourself where possible.
- Scan the archive with your antivirus, and check any published hashes against what you downloaded.
- Stay offline for the first test. Modified executables can trip anti-cheat systems and may break online play. Rockstar has not endorsed this project.
Most guides skip the last point. Running a modified game executable can put your online account at risk, and that's a bigger cost than a few FPS.
Known Issues Worth Knowing About
The README lists three known issues. The first launch needs extra time to build the shader cache. Some third-party overlays, such as performance monitors and recording tools, can interfere with the framework's measurements. Ultra-wide resolutions above 32:9 are labeled experimental.
The overlay issue deserves attention if you're testing. If you run an FPS counter that hooks into the render pipeline, you may measure the overlay's effect as well as the mod's. Test with one monitoring tool, and pick one you can reproduce results with.
What This Project Is Not
The repo's disclaimer is clear: this is an independent community effort, not affiliated with, endorsed by, or associated with Rockstar Games or Take-Two Interactive. Grand Theft Auto is a trademark of its owners. That distinction matters. A community project can't tell you what Rockstar plans for PC, what the official system requirements will be, or whether the game will support mods.
For platform and release details, check Rockstar Games' official site directly. Treat any third-party PC port as an experiment, not a substitute for an official build.
Frequently Asked Questions
What is the GTA 6 PC port on GitHub?
The GTA 6 PC port on GitHub, gta6recomp, is an experimental community framework. It aims to reduce shader compilation stutter, improve texture and asset streaming, and stabilize frame pacing. It is not made by Rockstar Games, and its performance claims come from an internal benchmark the author has not independently verified.
How do I install the GTA 6 PC port safely?
Back up your game folder first, then review the source code and scan the archive before running Setup.exe. Test offline, because modified executables can conflict with anti-cheat and put online accounts at risk. Keep the original install available so you can revert quickly if something breaks.
Why does GTA 6 stutter on PC the first time I play?
Stutter often happens because the GPU needs a compiled pipeline state object the first time it draws a new effect. The driver compiles it on the render path, which stalls the frame. Precompiled shader caches move this work earlier, but the cache itself takes time to build, so the first session can look worse until it's warm.
Can the gta6recomp mod improve FPS on an RTX 3060?
The README reports a jump from 58 to 71 average FPS on an RTX 3060 in one dense urban test scene. That's roughly a 22% gain on the author's numbers. Your results will depend on resolution, upscaler settings, driver version, and scene. Test with frame generation off, and compare frame-time consistency, not only average FPS.
Final Takeaway
The most useful thing you can do is treat gta6recomp as a hypothesis. If you want to test whether shader precompilation fixes your stutter, back up your install, run the mod offline, let the cache warm up, and measure 1% lows over a fixed route in a crowded area. Try that this week and note the frame-time graph before and after. Anyone wrestling with GTA 6 PC port stutter should start with that test before trusting any headline FPS number.