TinkerLab
ARTIFACT // LAUNCH-SP7-M5-ULTRA-FRONTIER-MOE.SH (SH)

EPYC SP7 (16-Ch MRDIMM) & Mac Studio M5 Ultra Frontier MoE Launcher (DeepSeek-V4.1-Flash / GLM-5.3 / Muse Spark) (launch-sp7-m5-ultra-frontier-moe.sh)

Audits 16-channel SP7 RDIMM/MRDIMM population and disables auto-NUMA on Linux, or raises Metal wired limit to 480GB on 512GB Mac Studio M5 Ultra before launching 552B–816B MoE inference.

[KEY VERIFIED TAKEAWAY / TL;DR]This reproducible configuration artifact "EPYC SP7 (16-Ch MRDIMM) & Mac Studio M5 Ultra Frontier MoE Launcher (DeepSeek-V4.1-Flash / GLM-5.3 / Muse Spark)" (launch-sp7-m5-ultra-frontier-moe.sh, SH) is paired with Tinker AI Lab's hardware-verified guide "amd-epyc-venice-sp7-16ch-mrdimm-vs-mac-studio-m5-ultra-deepseek-v41-flash-glm53-muse-spark". It exposes 4 customizable browser variables (HOST_TARGET=epyc_sp7, MODEL_PATH=/nvme/models/DeepSeek-V4.1-Flash-Q4_K_M.gguf, CTX_TOKENS=131072, SERVER_PORT=8080) with loopback-first security defaults.
01 // HOMELABLocal AI / HomeLab· Verified: 2026-11-17
Browse All 60 Artifacts →
launch-sp7-m5-ultra-frontier-moe.sh

EPYC SP7 (16-Ch MRDIMM) & Mac Studio M5 Ultra Frontier MoE Launcher (DeepSeek-V4.1-Flash / GLM-5.3 / Muse Spark)

Audits 16-channel SP7 RDIMM/MRDIMM population and disables auto-NUMA on Linux, or raises Metal wired limit to 480GB on 512GB Mac Studio M5 Ultra before launching 552B–816B MoE inference.

View Paired Guide →
Customizable Variables (Saved in Browser)
UTF-8 Encoded · Live Client-Side Compilation
#!/usr/bin/env bash
# Tinker AI — AMD EPYC SP7 (16-Ch MRDIMM) & Mac Studio M5 Ultra Frontier MoE Launcher (Guide #53)
set -euo pipefail

TARGET="epyc_sp7"
MODEL="/nvme/models/DeepSeek-V4.1-Flash-Q4_K_M.gguf"
CTX="131072"
PORT="8080"

if [ "$TARGET" = "epyc_sp7" ]; then
  echo "=== [EPYC SP7 16-Ch Mode] Auditing DDR5/MRDIMM Speed & Disabling Auto-NUMA ==="
  sudo dmidecode -t memory | grep -E "Speed:|Configured Memory Speed:" | sort -u || true
  sudo sysctl -w kernel.numa_balancing=0
  exec numactl --interleave=all llama-server -m "$MODEL" -t 96 -fa on --cache-type-k fp8_e4m3 -c "$CTX" --port "$PORT"
elif [ "$TARGET" = "mac_m5_ultra" ]; then
  echo "=== [Mac Studio M5 Ultra 512GB Mode] Unlocking 480GB Metal Wired Pool ==="
  sudo sysctl -w iogpu.wired_limit_mb=491520
  exec mlx_lm.server --model "$MODEL" --port "$PORT"
else
  echo "Unknown HOST_TARGET=$TARGET (expected: epyc_sp7 | mac_m5_ultra)" >&2
  exit 1
fi
PARAMETER SPECIFICATION // 参数化变量规范说明表

Customizable Parameter Reference for launch-sp7-m5-ultra-frontier-moe.sh

This SH artifact exposes 4 template variables compiled client-side in your browser.

Template TokenParameter DescriptionVerified DefaultExample Format
{{HOST_TARGET}}Host Mode (epyc_sp7 | mac_m5_ultra)epyc_sp7epyc_sp7
{{MODEL_PATH}}GGUF or MLX Model Path / Repo/nvme/models/DeepSeek-V4.1-Flash-Q4_K_M.gguf/nvme/models/DeepSeek-V4.1-Flash-Q4_K_M.gguf
{{CTX_TOKENS}}Context Window (-c)131072131072
{{SERVER_PORT}}OpenAI API Port80808080
PAIRED HARDWARE MANIFEST // 配套实测环境与边界记录

Verified Environment for "AMD 6th Gen EPYC "Venice" SP7 (16-Ch MRDIMM · 1.6 TB/s) & Helios vs Mac Studio M5 Ultra (512GB): Running DeepSeek-V4.1-Flash (552B), GLM-5.3 (744B) & Muse Spark Open (816B)"

Tested Hardware & OS

AMD 6th Gen EPYC "Venice" SP7 (16-Ch DDR5-8000 / MRDIMM-12800, 1TB RAM) & Helios Open Rack Blueprint vs Apple Mac Studio M5 Ultra (512GB Unified @ 1.2 TB/s)

Pinned Stack & Model Versions
llama.cpp v1.0.0+ik_llama.cpp (2026.11)vLLM 0.12+mlx-lm 0.29+Ubuntu 26.04 / macOS 26
Offline Boundary (100% Air-Gapped Local / On-Premise Frontier MoE)

DeepSeek-V4.1-Flash uses the permissive MIT license; GLM-5.3 and Meta Muse Spark Open use custom permissive licenses with commercial revenue/MaaS review thresholds—verify compliance before hosting public APIs.

Known Operational Limits & Failure Modes
  • Populating 16-channel SP7 with JEDEC MRDIMM-12800 (1.6 TB/s/socket) requires 1 DPC (16 DIMMs/socket) and compatible SP7 motherboards; mixing RDIMM speeds downclocks the entire memory controller.
  • Meta Muse Spark Open (816B total) weighs ~470–495 GB in 4-bit Q4_K_M—exceeding the ~475GB safe Metal wired pool on a 512GB Mac Studio M5 Ultra at long contexts unless quantized to 3.5-bit/IQ3_XS or run on 1TB+ SP7 EPYC.
  • Mac Studio M5 Ultra runs whisper-quiet under ~220W for 1–4 concurrent users, whereas an EPYC SP7 / Helios node draws 500W–1,200W+ and requires server rack airflow.
DEPLOYMENT CHECKLIST // 配置落地核验步骤

3-Step Verification Checklist After Deploying launch-sp7-m5-ultra-frontier-moe.sh

  1. Verified 16 populated DIMMs (1 DPC) per SP7 socket at rated 8,000 MT/s RDIMM or 12,800 MT/s MRDIMM speed via dmidecode.
  2. Disabled kernel.numa_balancing=0 and launched llama-server with numactl --interleave=all to prevent cross-CCD page thrashing.
  3. Raised iogpu.wired_limit_mb=491520 on 512GB Mac Studio M5 Ultra and verified MIT vs custom MaaS license compliance for DeepSeek-V4.1-Flash / GLM-5.3 / Muse Spark Open.