Repositoryashhart/MCDMATool
Running massive AI models locally often demands pooling memory across multiple workstations. Standard networking stacks introduce latency bottlenecks. MCDMA allows engineers to link high-end desktop machines directly via Thunderbolt or dedicated NICs, enabling low-latency model inference across mixed hardware.
MCDMA is in an active, early development phase (CLI 1.2.0, driver 0.1.18). Dual-Mac Thunderbolt bonding and Spark-to-Strix transfers are hardware-verified, though Strix-to-Mac setups still require physical confirmation. It represents a practical, niche solution for developers looking to scale LLM memory footprints across mixed workstation setups.
Summary generated by AI (Gemini 3.8 Flash, 7 October 2026). It may contain errors – the original source is authoritative.
Similar in content, found by the AI search.
X post:Ash Hart
Developer Ash Hart has announced bonded pair Thunderbolt 5 connectivity within the MCDMA RDMA framework. Paired with the TensorFold inference engine across two Apple Mac Studios running M5 Ultra chips, the dual-cable link delivers an 83 percent throughput uplift for tensor and pipeline parallelism.
216bookmarks24.3kviews
Repository:ashhart/TensorFold
TensorFold is an open-source inference engine designed to serve large language models on Apple Silicon (via MLX/Metal) and newer NVIDIA GPUs (Compute Capability 8.9+). It provides OpenAI- and Anthropic-compatible endpoints with model-tailored kernels and speculative draft verification.
1.1kstarsPython
Link:Apple
Apple has introduced the updated Mac Studio featuring M5 Max and M5 Ultra processors, focusing heavily on demanding creative, software engineering, and local AI workloads.
VideoVideo:Heavy Metal Cloud
Heavy Metal Cloud reviews the Gigabyte and ASUS Nvidia DGX Spark desktop units powered by the Grace Blackwell architecture. The video provides a step-by-step guide for local LLM serving using Docker, vLLM, and SparkRun, as well as media generation in ComfyUI, comparing real-world performance against a Mac Studio.
Repository:eugr/spark-vllm-docker
spark-vllm-docker is an open-source tool providing Docker configurations and startup scripts to run vLLM on DGX Spark systems, supporting both single-node setups and multi-node clusters via Ray or PyTorch distributed mode.
2.2kstarsShell
Link:Donato Capitella
Donato Capitella provides Strix Halo AI Toolboxes, a collection of containerised environments and host configuration guides for AMD's Ryzen AI MAX+ platform. Leveraging unified memory, the hardware allows the integrated GPU to address up to 128 GB of RAM for local LLMs, image generation, and model training.
Parts of this content were created with the assistance of AI.
