AI GlossaryCWords you meet while using AI
ConnectX-7
An ultra-fast NVIDIA network controller that lets multiple devices, like DGX Spark units, be cabled together and used as one.
In plain words
ConnectX-7 is a communication component that lets multiple computers exchange data quickly over cables. Think of a home router that links several rooms to the internet, except this one connects several small, single-room-sized computers at top speed so they can act like one large machine.
The two ports on the back of NVIDIA's compact AI computer, DGX Spark, are wired directly to this component. Plugging cables into those ports lets multiple DGX Spark units split a single job between them, which matters when a task is too big to fit on one machine and needs to be divided across several.
An unusual detail is that even with just one cable plugged into one port, the inside of the machine sees the connection as split into multiple channels. That's because the component internally links the external port to the computer's brain through two separate paths, so what looks like a single cable on the outside actually opens several communication lanes at once.
How it shows up in the news
The article explains that "the back of the DGX Spark has two QSFP (ConnectX-7) ports each capable of up to 200Gb/s." Here, QSFP is the name of the connector standard the cable plugs into, while ConnectX-7 is the name of the chip inside that port that actually handles the data exchange. The two are easy to confuse, but it helps to remember that QSFP is the outer shape and ConnectX-7 is the component inside it.
See also
Stories using this term
- NVIDIA Confirms Official Guide for Connecting Two, Three, and Four DGX Spark UnitsAI · 2026.09.05
- NVIDIA shares guide on running local AI by chaining multiple DGX Spark unitsAI · 2026.08.09
- Factory Builds AI Dev Environment Where Code Never Leaves the Machine, on DGX SparkAI · 2026.08.12
- NVIDIA RTX Spark laptops and mini PCs debut live at IFAAI · 2026.09.03
- NVIDIA unveils 'Ising Calibration 1.5' VLM for automated quantum computer calibrationAI · 2026.08.09
- Liquid AI's 300M Draft Model Speeds Up Decoding by Up to 3.18xAI · 2026.08.21
