unified-memory

Evidence-backed AMD Strix Halo local-AI setup and benchmarks: Qwen3.8, Ollama, llama.cpp, Vulkan/ROCm, large GGUFs, and cross-OEM results.

Honest local LLM deployment planning and benchmarking for high unified-memory Macs and future Linux/NVIDIA rigs.

Windows port of GreenBoost - GPU VRAM extension with system RAM & NVMe for larger LLMs. Original by Ferran Duarri (GPL v2).

Operator-grade GPU monitor for NVIDIA GPUs with native GB10 / DGX Spark coherent UMA support — PSI pressure, clock detection, ConnectX-7 network layer

Exact LLM inference engine written from scratch for AMD Strix Halo (gfx1151, ROCm/HIP). Qwen3.8-Flash-Next on one node or two replicas; DeepSeek V4.1 Flash across two nodes over USB4 with an optional RTX PRO. Speculative decoding bit-identical to greedy, up to 255 tok/s on coding chat. MiniMax and GLM next.

Local inference server for Apple Silicon — hot-swaps MLX models (LLM, vision, embeddings, TTS, STT) via OpenAI API

Apache Arrow compute on Apple silicon GPUs through Metal: Arrow columns in unified memory, copy-free where the producer's buffers are page aligned; null-aware GPU kernels; Arrow C Data, C Device and C Stream interop

A CUDA implementation of the transpose-free Quasi-Minimal Residual method

Unified Memory Abstraction Layer for AI Inference on AMD APUs and Intel iGPUs

A turnkey, fully-local AI workstation engineered for the AMD Ryzen AI Max+ 395. LLM inference, voice, document parsing, browser automation, agents — all on-device.

gpu thrashingNVIDIA GPU Unified Memory diagnostic tool — architecture-aware, measurement-based, PCIe/coherent transport detection

Talos-O (Omni): A sovereign, embodied agentic organism forged on AMD Strix Halo. Integrating the Chimera Kernel (Linux 7.0), Zero-Copy Introspection, and the Phronesis Engine. Built from First Principles.

SUPERFAST: turn an AMD Ryzen AI Max+ 395 (Strix Halo, gfx1151) into a fast, private local-LLM machine. Four profiles at maximum context: Qwen3.8-27B dense, Qwen3.8-Flash-Next MoE, Gemma-4-26B, DeepSeek-V4-Flash. Fedora 44 setup script, verified downloads, one-command switch, agent-ready API.

Fundamentals of Accelerated Computing C/C++ is a course provided by NVIDIA.

Research into CUDA Unified Memory as a VRAM extension for LLM inference

Empirical kernel scheduling characterization for NVIDIA GB10 (SM121a). Sweeps GEMM tile configurations, classifies PTX instruction paths, captures hardware telemetry

Measured llama.cpp Vulkan tuning for AMD Strix Halo (Ryzen AI MAX+ 395 / gfx1151), plus an eval harness with a catalogue of 13 bugs that each produced a believable wrong number.

NVML unified memory shim for NVIDIA DGX Spark Grace Blackwell GB10 - enables MAX Engine, PyTorch, and GPU monitoring

Run LLMs larger than your RAM — native GGUF inference engine with SSD streaming, no GPU required

System monitor that reports real GPU memory on unified-memory AMD parts (Strix Halo / Ryzen AI Max), where other tools show only the firmware carve-out. Split collector daemon + TUI, no dependencies.