spark-smi 3.6.0

A terminal-based system monitor (TUI) for NVIDIA Grace Blackwell (GB10) and hybrid CPU architectures

SPARK-SMI

A specialized terminal-based system monitor (TUI) built for NVIDIA Grace Blackwell (GB10) and hybrid ARM architectures — because nvidia-smi alone doesn't tell the full story.

---

Demo

---

Why SPARK-SMI?

The NVIDIA DGX Spark (GB10) is a unique system — a Grace Blackwell chip with unified CPU+GPU memory, hybrid Cortex-X925/A725 core clusters, and high-speed MT2910 200G networking. Standard tools like nvidia-smi, htop, and nvtop were not built with this topology in mind. SPARK-SMI was.

What it handles correctly	Standard tools
Hybrid P-core / E-core CPU clusters	❌
GB10 Unified Memory (CPU+GPU shared)	❌
MT2910 200G NIC bandwidth monitoring	❌
Mixed GPU architectures in one system	❌
NVML with graceful CLI fallback	✅
Zero system dependencies	✅

---

Features

• Snapshot Mode — Runs once and prints to stdout with full ANSI colors, just like nvidia-smi. Pipe it, log it, script it.
• Live Mode (-l) — Flicker-free TUI that refreshes every second with responsive terminal resize handling.
• Hybrid CPU Topology — Correctly splits and labels Cortex-X925 (Performance) and Cortex-A725 (Efficiency) core clusters with individual per-core load bars.
• Unified Memory Aware — Detects GB10 unified memory architecture and maps system RAM to GPU memory display accurately.
• Dual GPU Support — Handles mixed architectures (e.g. sm_121 GB10 + sm_86 RTX 3090 via OcuLink) simultaneously.
• NIC Monitoring — Real-time bandwidth utilization across all interfaces: MT2910 200G ports (1–4) and Realtek 10G (port 5), read directly from sysfs.
• Driver & CUDA Info — Footer displays live Driver version and CUDA version via NVML or nvidia-smi fallback.
• Robust Fallbacks — NVML → nvidia-smi CLI → graceful degradation. Adapts to missing sensors, fan controllers, and unsupported queries without crashing.

---

Screenshots

Full Dashboard	Resize-Safe
	Scales cleanly from narrow to full-width

---

Prerequisites

• Linux (aarch64 recommended — built and tested on DGX Spark)
• Python 3.6+
• NVIDIA Drivers installed
• nvidia-smi in PATH

---

Installation

Option 1: Quick Run (Virtual Environment)

The safest method on DGX appliances — no system libraries touched.

git clone https://github.com/chappa-ai-llc/spark-smi.git
cd spark-smi
python3 -m venv venv
./venv/bin/pip install -r requirements.txt

# Snapshot (single output)
./venv/bin/python3 spark-smi.py

# Live mode
./venv/bin/python3 spark-smi.py -l

Option 2: System Alias (Recommended)

Type spark-smi from anywhere.

echo "alias spark-smi='~/spark-smi/venv/bin/python3 ~/spark-smi/spark-smi.py'" >> ~/.bashrc
source ~/.bashrc

---

Usage

Command	Action
spark-smi	Snapshot — print once and exit
spark-smi -l	Live mode — interactive TUI
spark-smi -n 0.5 -l	Live mode at 0.5s refresh rate

Interactive Controls

Key	Action
q	Quit
t	Toggle temperature units (°C / °F)
u	Toggle memory units (GiB / GB)

---

Tested Hardware

Component	Details
System	NVIDIA DGX Spark
SoC	GB10 Grace Blackwell (sm_121)
External GPU	RTX 3090 via M.2 OcuLink (sm_86) — mixed architecture, single dashboard
NICs	MT2910 × 4 (200G, 100G & 40G DAC), Realtek × 1 (10G, 5G, 2.5G, 1G)
OS	Linux 6.17.0-nvidia
Driver	580.126.09
CUDA	13.0

---

Roadmap

• Fan Monitoring — Read GB10 chassis fan speeds without sudo (currently blocked by nvsm/IPMI privilege requirements)
• REST API / Prometheus Exporter — Expose a lightweight JSON HTTP endpoint for Grafana and Prometheus integration
• CSV Logging Mode — --csv flag to pipe raw metrics to stdout or file for external processing
• PyPI Package — pip install spark-smi one-liner install
• Multi-node Support — Monitor clustered DGX Spark nodes from a single dashboard

---

About

Built by chappa-ai-llc — a solo homelab project born out of frustration with existing tools on novel hardware.

If this saved you time, a ⭐ on the repo is appreciated.

---

License

MIT — see LICENSE for details.