Open-Source Systems Software · C++20 · Windows x64 · MIT
Run DirectX 12 (DX12) on NVIDIA GT 740 & Kepler GPUs
KeplerBridge (github.com/HamzaShaikhAI/KeplerBridge) is a drop-in 64-bit d3d12.dll compatibility layer that translates Direct3D 12 API calls and Shader Model 6.0 DXIL bitcode onto native Direct3D 11 Feature Level 11_0 (0xB000) NVIDIA Kepler GPUs—requiring zero driver patches and zero Windows system-file changes.
~1.12 ms → ~0.03 msWarm SHA-256 disk shader cache (>35x fixture speedup)
86 TitlesCatalogued in docs/GAME_COMPATIBILITY.md
FL 11_0 (0xB000)Model A honest Kepler capability reporting
0 System EditsUser-space d3d12.dll beside the game .exe
What is KeplerBridge and why do GT 740 GPUs crash on DX12?
KeplerBridge is an open-source ($0, MIT-licensed) C++20 Direct3D 12-to-Direct3D 11 translation layer authored by Hamza Shaikh (@HamzaShaikhAI) in 2026 to unlock compatible DX12 workloads on legacy NVIDIA Kepler cards. According to Microsoft Direct3D feature documentation and NVIDIA driver releases, Kepler GPUs—such as the GeForce GT 740 (GK107/GK208), GTX 750, GTX 760, GTX 770, GTX 780, and GTX TITAN—are capped at Feature Level 11_0 (0xB000), causing DX12 titles using Shader Model 6.0 DXIL to crash with DXGI_ERROR_UNSUPPORTED (0x887A0004).
In practice, KeplerBridge solves startup rejection in three user-space steps. Step 1: drop 64-bit d3d12.dll beside the game executable with 0% Windows system modification. Step 2: run the standalone LLVM 3.7 decoder to lower DXIL shaders into SM 5.0 DXBC bytecode. Step 3: serialize draw calls onto Direct3D 11 using Model A capability contracts across 86 catalogued PC titles.
How fast is the DXIL-to-DXBC translator and shader cache?
According to public KeplerBridge v0.5.0-rc1 benchmarks (November 2026), warm SHA-256 disk shader cache loading reduces shader startup overhead by 97.3%, dropping cold DXIL-to-DXBC translation latency from 1.12 ms down to 0.03 ms (over 35x faster) on synthetic test fixtures. For example, warm cache hits skip repeated LLVM 3.7 bitstream parsing during subsequent game launches.
First, straight-line Mem2Reg SSA promotion lowers scalar stack allocations into virtual registers without linking external Clang libraries. Second, synthetic critical-edge splitting and bounded natural loop lowering guarantee static termination across shader stages using 100% deterministic hashing. Third, compute shaders missing native Kepler wave intrinsics (such as WaveActiveSum) use fast parallel reduction trees in groupshared memory. Note that per-game FPS projections across all 86 titles are engineering estimates for a 4 GB GT 740 rather than universal guarantees.
Step-by-Step Setup
Drop-In Installation (Windows x64)
User-space proxy deployment requires zero Windows system-directory changes and zero NVIDIA driver modifications.
Step 01 · Obtain
Release Binary or Source Build
Download KeplerBridge-0.5.0-rc1-Windows-x64.zip from the GitHub Release page, or compile locally using CMake and MSVC with C++20 enabled.
Step 02 · Deploy
Place Next to the Game .exe
Extract d3d12.dll and kepler_bridge.ini directly into the folder containing the 64-bit game executable. Never place proxy DLLs inside C:\Windows\System32.
Step 03 · Verify
Run & Check Diagnostics
Launch in offline single-player mode. Never inject into kernel anti-cheat multiplayer games. Inspect kepler_bridge.log next to the executable to verify DXIL translation and Feature Level 11_0 execution.
CMake Build & Verification (PowerShell)
git clone https://github.com/HamzaShaikhAI/KeplerBridge.git
cd KeplerBridge
cmake -B build -DCMAKE_BUILD_TYPE=Release
cmake --build build --config Release
# Copy build\Release\d3d12.dll & kepler_bridge.ini next to your offline DX12 game .exe
Dynamic root constant tracking, texture footprint clamping, SHA-256 disk cache.
34–46 FPS
Playable on KeplerBridge (Estimate)
Lies of P / Hi-Fi RUSH
Unreal Engine 4.27
Linear root signature mapping, discard swapchain normalization, warm cache hits.
30–55 FPS
Playable on KeplerBridge (Estimate)
Hades II
Custom Engine (D3D12)
2D isometric draw path with lightweight constant buffers and low compute pressure.
50–60 FPS
Playable on KeplerBridge (Estimate)
Palworld
Unreal Engine 5.1
Requires D3D12 Tier 3 bindless descriptors absent on Kepler; launch using engine flag -dx11 instead.
30–45 FPS (-dx11)
D3D12 Unsupported (Use Native -dx11)
Disclaimer: FPS figures in the 86-game compatibility showcase are engineering estimates for a GT 740 (4 GB DDR3) at 720p Low presets unless explicitly labeled as benchmark measurements. Online anti-cheat multiplayer titles are unsupported.
Direct Reference
KeplerBridge FAQ
Can KeplerBridge run DirectX 12 games on an NVIDIA GeForce GT 740?
Yes, for supported workloads. KeplerBridge (github.com/HamzaShaikhAI/KeplerBridge, release candidate v0.5.0-rc1) is an open-source C++20 Direct3D 12-to-Direct3D 11 compatibility proxy engineered specifically for legacy NVIDIA Kepler GPUs including the GeForce GT 740 (GK107/GK208), GTX 750, GTX 760, GTX 770, GTX 780, and GTX TITAN at Feature Level 11_0.
Deployed in user space as a drop-in 64-bit d3d12.dll beside the game executable with 0 Windows system modifications, KeplerBridge translates Shader Model 6.0 DXIL into Shader Model 5.0 DXBC using a cleanroom LLVM 3.7 decoder and SHA-256 disk shader cache (~1.12 ms cold to ~0.03 ms warm on synthetic fixtures). Compatibility is catalogued across 86 PC titles and remains workload-dependent.
Does KeplerBridge work on Unreal Engine 5 games requiring Tier 3 bindless or ray tracing?
No. Titles that strictly mandate Direct3D 12 Tier 3 bindless descriptor indexing, mesh shaders, or hardware ray tracing exceed NVIDIA Kepler silicon limits and cannot be translated onto Feature Level 11_0 hardware. Unreal Engine 5.1 titles such as Palworld crash in D3D12 on Kepler and should instead be launched with the game engine's native -dx11 switch when provided.