Nvidia RTX Spark Brings Local AI Agents to Personal PCs
Nvidia detailed RTX Spark in Seoul, a single-chip GPU and CPU platform built to run AI agents directly on a PC without a cloud connection.
SUMO· the number puzzle
Dialo· learn languages by speaking
ediviewer· read EDI files like documents
giftdetect· the right gift in minutesNvidia is pushing into the local AI market with RTX Spark, a superchip that combines a GPU and a CPU in one package so that AI assistants can run on a personal PC without relying on the cloud. The company presented the product's core architecture and local agent use cases at a press briefing in Seoul on November 6.
RTX Spark, also referred to as N1X, pairs an Nvidia Blackwell RTX GPU with a 20-core Grace CPU co-designed with MediaTek. The two chips are linked through NVLink-C2C, Nvidia's high-speed chip-to-chip interconnect, and the platform supports up to 128GB of unified memory and 1 petaflop of AI compute. The product was first introduced to Korea by CEO Jensen Huang in June and went on sale in the country after a launch event at Coex in Seoul on October 16.
One memory pool for CPU and GPU
On conventional laptops, graphics memory is limited by the class of the GPU. RTX Spark instead lets the CPU and GPU share the same memory space and keeps their caches coherent. According to Nvidia, that allows large AI models and demanding creative work to run side by side in a single pool of memory. The familiar GeForce RTX software stack, including CUDA, DLSS and TensorRT, is supported as is.
Nvidia says RTX Spark delivered 2.7 to 4.6 times the performance of typical systems-on-chip in tasks such as running local agents, fine-tuning, 3D rendering, image and video generation, and gaming. Jeff Yen, Nvidia's technical marketing director for Asia-Pacific, said no Windows SoC comes close in performance and power efficiency.
Why agents, and why now
Yen called this year an inflection point for local agents and argued that their intelligence is quickly catching up with cloud-based frontier models. Unlike conventional LLM chatbots, he said, AI agents understand a user's goal, reason about it and carry out actions, turning the PC into something closer to a teammate than a tool.
Alongside the hardware, Nvidia showed lightweight models that sharply cut memory use, as well as technology that mixes local and cloud processing. The company's stated aim is to complete an on-device AI ecosystem around the chip.
Apps worth a look
SUMO: the number puzzleTap the right tiles, add them up, hit the target. Easy to learn, hard to master. Free, for all ages.
Dialo: learn languages by speakingTalk to an AI conversation partner; every sentence comes back corrected and explained in your language. 9 languages, A1 to C2.
ediviewer: read EDI files like documentsOpen EDIFACT and X12 messages, see every segment explained, convert to JSON/XML, export PDF and Excel. Runs locally, no upload.
giftdetect: the right gift in minutesPick who it is for, the occasion and their personality. The search engine reads product descriptions from local shops and ranks what fits. Free, no sign-up.
More from this day
See all →
Citus Demonstrates GPS-Free Firefighter Tracking, Eyes Military Use
Google Cloud expands CodeMender and AI agent security tools
Operation Ghostbusters: 830,840 Suspected Ghost Students Found
