MiniMax-H3 Optimisation for NVIDIA DGX Spark (sm_121)

Repositorydrowzeys/keys-SM121-Optimized-MiniMax-H3-Nvidia-Sol-Engine-Kijai-SolAttn_Triton-Single-DGX-SparkTool

An optimised ComfyUI setup for the MiniMax-H3 (33B) audio+video model on a single NVIDIA DGX Spark (GB10, sm_121). It fixes a silent miscompilation issue and accelerates end-to-end rendering by 1.54x.

Key points

  1. Fixes silent miscompilation on sm_121 where Sol-Attn defaulted to flex_attention, silently generating broken outputs without warnings.
  2. Ports NVIDIA Sol-Engine features (FirstBlockCache and batched VAE tile decoding) from multi-GPU SGLang to single-GPU ComfyUI.
  3. Reduces rendering time for a 5-second 864x480 clip from 312.7 s to 202.5 s (1.54x end-to-end, 1.72x per denoising step) with no visible quality loss.
  4. Integrates kijai's Triton INT8/TMA kernels, including an automated installer script (install.sh) and an agent recipe.
  5. Validates that multi-node distribution (CP=2) is slower on DGX Spark due to interconnect bandwidth limits (5.52 GB/s vs 503 GB/s NVLink) compared to a tuned single node.

Why it matters

New architectures like NVIDIA DGX Spark (GB10 / sm_121) can suffer from silent kernel compilation bugs that waste significant debugging time. This project fixes these issues for MiniMax-H3 and demonstrates how enterprise multi-GPU optimisations can be backported to local ComfyUI environments.

Analysis

The project is tailored specifically for NVIDIA DGX Spark (GB10) hardware and explicitly blocks other chips; RTX 5090 users are advised to use unpatched upstream plugins. In terms of maturity, it is a focused, open-source utility repository (30 stars, Apache 2.0) featuring reproducible benchmark data, targeted install automation, and a narrow hardware scope.

Mentioned links

Found in

Summary generated by AI (Gemini 3.8 Flash, 27 September 2026). It may contain errors – the original source is authoritative.

Related entries

Similar in content, found by the AI search.

  • X post:keys

    Accelerating MiniMax-H3 on Nvidia DGX Spark with ComfyUI and Sol-Engine

    A public repository demonstrates a 1.54x end-to-end acceleration for the 33B MiniMax-H3 omni-DiT on a single Nvidia DGX Spark (GB10, sm_121) using ComfyUI. It ports Nvidia Sol-Engine techniques, applies Triton patches, and resolves critical Blackwell compilation issues.

    110bookmarks15.6kviews

    AI & LLMs· Tool

  • Repository:MiaAI-Lab/DeepSeek-v4-Flash-One-DGX-Spark

    Running DeepSeek-v4-Flash on a Single DGX Spark

    An open-source launcher and Docker recipe for running DeepSeek V4 Flash 0731 (EXL3) on a single NVIDIA DGX Spark with 128 GiB unified memory.

    328starsPython

    AI & LLMs· Tool

  • Repository:MiaAI-Lab/GLM-5.3-Flash-EXL3-2x-DGX-Sparks

    Serving GLM-5.3 Flash EXL3 on Dual NVIDIA DGX Spark Kits

    Mia's AI Lab provides an optimised vLLM deployment overlay for GLM-5.3-Flash using 4 bpw EXL3 quantisation tailored for 2x NVIDIA GB10 systems.

    431starsPython

    AI & LLMs· Tool

  • Link:NVIDIA

    NVIDIA DGX Spark Developer Hub: Workflows for GB10 Grace Blackwell Superchip

    NVIDIA has launched the DGX Spark Developer Hub, offering playbooks and guides for running AI agent workflows on the NVIDIA GB10 Grace Blackwell Superchip. The hub features quickstarts for tools such as Open-WebUI, ComfyUI, and vLLM alongside multi-Spark networking setups.

    AI & LLMs· Resource

  • Video

    Video:Heavy Metal Cloud

    Nvidia DGX Spark Hands-On: Setup, vLLM Serving, and Mac Studio Comparison

    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.

    AI & LLMs· Tutorial

Back to the knowledge base

Let's talk about your project

Get in touch

Locations

  • Mattersburg
    Johann Nepomuk Bergerstraße 7/2/14
    7210 Mattersburg, Austria
  • Vienna
    Ungargasse 64-66/3/404
    1030 Wien, Austria

Parts of this content were created with the assistance of AI.