rtx-5080

RTX 5090 & RTX 5060 Docker container with PyTorch + TensorFlow. First fully-tested Blackwell GPU support for ML/AI. CUDA 12.8, Python 3.11, Ubuntu 24.04. Works with RTX 50-series (5090/5080/5070/5060) and RTX 40-series.

Selective fused FP8 checkpoint + ComfyUI runtime for LingBot-Video Dense 1.3B. Measured 1.27x in a local one-step RTX 5080 sampler smoke test; T2V, TI2V, experimental FLF.

Provide tested tools and configs to run Qwen 3.5 GGUF models efficiently on a single 16GB NVIDIA GPU using llama.cpp locally.

Downstream llama.cpp TurboQuant CUDA fork with adaptive KV layout selection for long-context inference on consumer Blackwell GPUs.

Kernel module unlocking the GPU power limit on HP OMEN MAX 16 (8D87) from 80 W to 175 W, with a fail-closed watchdog

CUDA backend for rmems/agoge-forger: local LLM fine-tuning and training kernels on RTX 5080 (Blackwell sm_120, ~16 GB).

Single-GPU passthrough of an RTX 50 series (Blackwell, 5080 in test case) to a VM on Linux: scripts, EndeavourOS ISO patcher, and the fixes for the black guest screen and the black host display after guest shutdown.

True Q4_K_M Qwen3.8-27B at 64K on an RTX 5080 16 GB using selective FFN offload, llama.cpp, and Pi.

Not a product. Not a framework. Nothing here is packaged for you to deploy. This is my house, my desk, my power bill, and the machine that thinks in it. It exists so I can point at something on my wall and say that box runs my world.

Pre-built onnxruntime-gpu 1.24.1 with Blackwell sm_120 CUDA kernels (RTX 5090/5080/5070)