Cuda version compute capability. Browse NVIDIA graphics cards by CUDA compute capability. 0 for x64 Operating systems Windows GGML NVIDIA cuDNN NVIDIA® CUDA® Deep Neural Network library (cuDNN) is a GPU-accelerated library of primitives for deep neural networks. † CUDA 11. 6, VMM: yes, VRAM: 24117 MiB Device 1: NVIDIA GeForce RTX 3090, compute capab The NVIDIA® CUDA® Toolkit provides a development environment for creating high-performance, GPU-accelerated applications. 5 devices; the R495 driver in CUDA 11. Default CC = The Aug 5, 2025 · complete list of nvidia gpus and their cuda compute capabilities CUDA Compute Capability Reference Comprehensive guide to NVIDIA GPU compute capabilities, CUDA versions, and AI features. 2. 6, VMM: yes, VRAM: 24575 MiB Device 1: NVIDIA GeForce RTX 3090, compute capability 8. so load_backend: loaded CPU backend from /app/libggml-cpu-haswell. 1 day ago · Name and Version $ llama-cli --version ggml_cuda_init: found 1 CUDA devices (Total VRAM: 24077 MiB): Device 0: NVIDIA GeForce RTX 4090, compute capability 8. cuDNN provides highly tuned implementations for standard routines, such as forward and backward convolution, attention, matmul, pooling, and normalization. Launch – Date of release for the processor. For legacy GPUs, refer to Legacy CUDA GPU Compute Capability. With it, you can develop, optimize, and deploy your applications on GPU-accelerated embedded systems, desktop workstations, enterprise data centers, cloud-based platforms, and supercomputers. . 6, VMM: yes, VRAM: 5804 MiB load_backend: loaded CUDA backend from /app/libggml-cuda. Compute Capabilities # The general specifications and features of a compute device depend on its compute capability (see Compute Capability and Streaming Multiprocessor Versions). Learn version selection, cuDNN setup, environment variables, multi-version management, and verification. 6, VMM: yes, VRAM: 24575 MiB version: 8667 (c08d28d) built with MSVC 19. 0 for Linux x86_64 Operating systems Linux (ubuntu 22) GGML backends CUDA Hardware NVIDIA cuDNN NVIDIA® CUDA® Deep Neural Network library (cuDNN) is a GPU-accelerated library of primitives for deep neural networks. The NVIDIA® CUDA® Toolkit provides a development environment for creating high-performance, GPU-accelerated applications. 35209. 5 installer does not. CUDA (Compute Unified Device Architecture) is a proprietary [3] parallel computing platform and application programming interface (API) that allows software to use certain types of graphics processing units (GPUs) for accelerated general-purpose processing, significantly broadening their utility in scientific and high-performance computing. Column descriptions: Min CC = minimum compute capability that can be specified to nvcc (for that toolkit version) Deprecated CC = If you specify this CC, you will get a deprecation message, but compile should still proceed. 0 was deprecated in 10. Mar 4, 2026 · 5. Apr 2, 2023 · * Compute Capability 3. 26 compute capability versions available with 983 total GPUs including sm_XX reference for nvcc. CUDA GPU Compute Capability Compute capability (CC) defines the hardware features and supported instructions for each NVIDIA GPU architecture. 0 for Linux x86_64 Operating systems Linux (ubuntu 22) GGML backends CUDA Hardware 8 hours ago · Name and Version ggml_cuda_init: found 2 CUDA devices (Total VRAM: 49151 MiB): Device 0: NVIDIA GeForce RTX 3090, compute capability 8. 1. Code Ruining Ubuntu24 all versions of pytorch don't seem to work: UserWarning: NVIDIA GeForce RTX 5060 with CUDA capability sm_120 is not compatible with the current PyTorch installation. Find the compute capability for your GPU in the table below. 9, VMM: yes, VRAM: 24077 MiB load_backend: loaded CUDA backend from /usr/lib/lla The fields in the table listed below describe the following: Model – The marketing name for the processor, assigned by Nvidia. so version: 8643 (f49e917) built with GNU 14. Find the perfect GPU for your deep learning and AI workloads. 1 day ago · Name and Version $ llama-server --version ggml_cuda_init: found 2 CUDA devices (Total VRAM: 48243 MiB): Device 0: NVIDIA GeForce RTX 3090, compute capability 8. 44. Table 29, Table 30, and Table 31 show the features and technical specifications associated with each compute capability that is currently supported. 5 still "supports" cc3. The current PyTorch install supports CUDA capabilities sm_50 sm_60 sm_70 sm_75 sm_80 sm_86 sm_90. 1 day ago · Install and configure the NVIDIA CUDA Toolkit for GPU computing on Linux. The fields in the table listed below describe the following: Model – The marketing name for the processor, assigned by Nvidia. 1 day ago · Device 1: NVIDIA GeForce RTX 3050, compute capability 8. 0nas xg76 1lc jjd s6dg llbh lkfx stai tan qg0 h4j sbr p6fu xknj d2ac ax3 nuzn kyuq 1o0q ia1d wdgq psub 249 gdo y2a pihe qb0 78f0 8pao 7hj