gfx1100

DEPRECATED!

在 AMD Radeon RX 7900 XTX（gfx1100 / RDNA3）上运行 DLSS Neural Rendering 的大力喜鹊分支 — DLSSNR served by an AMD backend

Native HIP/C++ Z-Image-Turbo inference optimized for AMD Radeon RX 7900 XTX gfx1100 WMMA.

Multi-GPU tensor-parallel vLLM on AMD Radeon RX 7900 XT / XTX / GRE (7900XT, 7900XTX, RX7900XT, gfx1100, RDNA3, ROCm): root cause and fix for the RCCL hostcall / PCIe atomics (AtomicOps) crash "NCCL error: unhandled cuda error" / "operation cannot be performed in the present state", Proxmox VFIO passthrough; LLM inference benchmarks on 13 machines

High-performance FP32 GEMV for AMD RDNA 3 (gfx1100), with reproducible performance and numerical analysis.

FlashAttention-2 Triton kernel for AMD RDNA3 (gfx1100 / Radeon RX 7900 XT/XTX/GRE): forward, backward, and split-K decode

Early Compatibility for RDNA 3 cards for ML/HIP

LLM serving benchmarks and multi-GPU engineering findings from a dual AMD W7800 (RDNA3/gfx1100) workstation — bake-off data, working TP=2 recipe, W4A16 field validation

Measurement notes and tools from a local LLM and diffusion stack on consumer AMD hardware. Pre-registered noise bands, deterministic evaluation, and null results reported as plainly as positive ones.

Running LLMs locally on AMD GPUs, by people who actually do it. Cookbook + benchmark database for ROCm/HIP inference on Radeon, Instinct, and Strix Halo.

Run SenseNova-U1.5-8B-MoT (unified multimodal understanding + image generation) on AMD RDNA3 GPUs via ROCm — evidence-first, one command, receipts for every claim

Enabling PyTorch on XLA Devices (e.g. Google TPU)

Yolo ROCm build - llm slop

⚡ Automated nightly builds & portable releases for lucebox (DFlash) inference server on NVIDIA CUDA (sm_75-120) and AMD ROCm 7 (Strix Halo gfx1151, RX 7900 XTX gfx1100, R9700 gfx1201) with zero-dependency $ORIGIN RPATH bundling.

Discrete AMD ROCm (gfx1100 / RDNA3) build & deploy recipes for antirez's ds4 (DeepSeek-V4): 4-GPU full residency ~22 t/s + a one-URL ds4/vLLM/ollama VRAM switcher

ROCm/HIP port of cuPDLP-C for AMD Radeon 890M / gfx1150 with cross-device CUDA/ROCm benchmarks.

Serve AMD's Quark W4A16 INT4 models on RDNA3 (gfx1100) via a lossless repack to SGLang's GPTQ path

Accelerate Z-Image-Turbo inference on AMD Radeon RX 7900 XTX with custom HIP/C++ optimizations for maximum performance.