fix!: require CUDA compute cap 8.0

README.md

@@ -140,7 +140,7 @@ Koharu bundles CUDA Toolkit 13.0. The required DLLs are extracted to the applica
 
 #### Supported NVIDIA GPUs
 
-Koharu supports NVIDIA GPUs with compute capability 7.5 or higher.
+Koharu supports NVIDIA GPUs with compute capability 8.0 or higher.
 
 For GPU compatibility references, see [CUDA GPU Compute Capability](https://developer.nvidia.com/cuda-gpus).
 

docs/en-US/explanation/acceleration-and-runtime.md

@@ -10,7 +10,7 @@ Koharu supports multiple runtime backends so the same pipeline can run across a
 
 CUDA is the primary GPU backend on systems with supported NVIDIA hardware.
 
-- Koharu supports NVIDIA GPUs with compute capability 7.5 or higher
+- Koharu supports NVIDIA GPUs with compute capability 8.0 or higher
 - Koharu bundles CUDA toolkit 13.0
 
 On first run, Koharu extracts the required dynamic libraries into the application data directory.

docs/ja-JP/explanation/acceleration-and-runtime.md

@@ -10,7 +10,7 @@ Koharu は複数のランタイム経路を備えており、幅広いハード
 
 CUDA は、対応する NVIDIA ハードウェアを持つ環境での主な GPU アクセラレーション経路です。
 
-- Koharu は compute capability 7.5 以上の NVIDIA GPU をサポートします
+- Koharu は compute capability 8.0 以上の NVIDIA GPU をサポートします
 - Koharu は CUDA Toolkit 13.0 を同梱しています
 
 初回実行時には、必要な動的ライブラリがアプリケーションデータディレクトリへ展開されます。

docs/pt-BR/explanation/acceleration-and-runtime.md

@@ -10,7 +10,7 @@ O Koharu suporta múltiplos backends de runtime para que a mesma pipeline possa
 
 CUDA é o backend principal de GPU em sistemas com hardware NVIDIA suportado.
 
-- O Koharu suporta GPUs NVIDIA com compute capability 7.5 ou superior
+- O Koharu suporta GPUs NVIDIA com compute capability 8.0 ou superior
 - O Koharu empacota o CUDA toolkit 13.0
 
 Na primeira execução, o Koharu extrai as bibliotecas dinâmicas necessárias para o diretório de dados do aplicativo.

docs/zh-CN/explanation/acceleration-and-runtime.md

@@ -10,7 +10,7 @@ Koharu 支持多条运行时路径，因此可以在不同类型的硬件上运
 
 CUDA 是在支持的 NVIDIA 硬件上最主要的 GPU 加速路径。
 
-- Koharu 支持计算能力 7.5 及以上的 NVIDIA GPU
+- Koharu 支持计算能力 8.0 及以上的 NVIDIA GPU
 - Koharu 内置 CUDA Toolkit 13.0
 
 首次运行时，所需的动态库会被解压到应用数据目录中。

koharu-runtime/src/cuda.rs

@@ -14,7 +14,7 @@ const CUDA_13_1_DRIVER_VERSION: i32 = 13010;
 const CUDA_EXTRACT_REVISION: u32 = 2;
 const CU_DEVICE_ATTRIBUTE_COMPUTE_CAPABILITY_MAJOR: i32 = 75;
 const CU_DEVICE_ATTRIBUTE_COMPUTE_CAPABILITY_MINOR: i32 = 76;
-const MIN_COMPUTE_CAPABILITY: (i32, i32) = (7, 5); // Turing (RTX 20xx) and above
+const MIN_COMPUTE_CAPABILITY: (i32, i32) = (8, 0); // Ampere (RTX 30xx) and above
 
 type CuInit = unsafe extern "C" fn(flags: u32) -> i32;
 type CuDriverGetVersion = unsafe extern "C" fn(driver_version: *mut i32) -> i32;
@@ -158,7 +158,7 @@ pub fn driver_version() -> Result<CudaDriverVersion> {
 
 /// Query the compute capability of CUDA device 0.
 ///
-/// Returns `(major, minor)` e.g. `(7, 5)` for Turing, `(8, 6)` for Ampere.
+/// Returns `(major, minor)` e.g. `(8, 0)` for Ampere, `(8, 9)` for Ada.
 pub fn compute_capability() -> Result<(i32, i32)> {
     let library_name = if cfg!(target_os = "windows") {
         "nvcuda.dll"
@@ -245,7 +245,7 @@ pub fn check_cuda_driver_support() -> bool {
         }
     }
 
-    // Check GPU compute capability (need >= 7.5 / Turing)
+    // Check GPU compute capability (need >= 8.0 / Ampere)
     match compute_capability() {
         Ok((major, minor)) if (major, minor) >= MIN_COMPUTE_CAPABILITY => {
             tracing::info!("GPU compute capability: {major}.{minor}");
@@ -254,7 +254,7 @@ pub fn check_cuda_driver_support() -> bool {
         Ok((major, minor)) => {
             tracing::warn!(
                 "GPU compute capability {major}.{minor} is below the minimum \
-                 required {}.{}; falling back to CPU. A Turing (RTX 20xx) or \
+                 required {}.{}; falling back to CPU. An Ampere (RTX 30xx) or \
                  newer GPU is required for GPU acceleration.",
                 MIN_COMPUTE_CAPABILITY.0,
                 MIN_COMPUTE_CAPABILITY.1,