Nvidia DGX Spark update cuts idle power by 32% or more — hot-plug detection on ConnectX NIC makes for a more efficient AI workstation

Source: Tom’s Hardware

Overview

Nvidia’s DGX Spark has proven itself as a compact, versatile local AI engine thanks to its 128 GB of RAM, fast 20‑core Arm CPU, and capable Blackwell GPU. It is equally useful for LLM inference and generative image and video workflows, as detailed in our review.

Idle Power Consumption

During testing, we observed that despite the DGX Spark’s 3 nm‑class fabrication technology, complex Arm CPU, and Nvidia’s extensive experience with mobile power management, the system drew a surprising ~37 W at idle. While not catastrophic for an otherwise efficient machine, this figure was higher than expected.

Fix

We discovered that enabling hot‑plug detection on the ConnectX NIC reduces idle power consumption by 32 % or more, providing a simple and effective way to make the AI workstation more power‑efficient.