Nvidia launches AI "super chip" for PCs.

Nvidia describes the RTX Spark as "the most efficient PC chip ever built," designed to run AI Agents locally rather than relying solely on cloud computing.

Announced on June 2nd at the Computex 2026 event in Taiwan, Nvidia representatives stated that the product is developed from Nvidia's graphics processor in collaboration with its two partners, MediaTek and Microsoft.

On its website, Nvidia introduces the RTX Spark as a "super chip," combining an Arm Grace CPU and a Blackwell RTX GPU on a single, unified architecture. The chip has up to 20 CPU cores, a Blackwell GPU with 6,144 CUDA cores, AI performance reaching 1 petaflop FP4, and supports up to 128 GB of LPDDR5X memory.

Illustration of the RTX Spark chip in a laptop. Photo: Nvidia

The chip is expected to be available in September, initially in about 30 laptop models and 10 desktop models from partners including Dell, Lenovo, Microsoft, HP, Asus, and MSI. According to Mark Aevermann, Nvidia's Director of Product Development, computers using this chip will "target content creators, AI developers, and gamers," and will be priced in the high-end segment.

Kari Briski, Vice President of AI Generation at Nvidia, emphasized that while GPUs have helped the company become the world's most valuable, agent computing is increasingly reliant on CPUs. According to him, AI agents will soon replace the use of AI in the form of chatbots like ChatGPT or Gemini.

"That era is coming to an end," Briski stated on stage at the event. "Agents will take on new workloads. They will operate everywhere, from data centers to endpoints."

The WSJ noted that RTX Spark received positive reviews on the stock market, with Nvidia shares rising 4%, Microsoft up 2.7%, while Apple fell 0.8%, and AMD, Intel, and Qualcomm declined between 4.9% and 8.5%.

"RTX Spark will create its own moment, just like the iPhone, ChatGPT, or DeepSeek did," Neil Shah, co-founder of Counterpoint Research, told Reuters. "RTX Spark aims to transform traditional application-focused PCs into truly useful personal AI systems. Ultimately, they will be in every home in the coming years, as private edge AI agents become more preferred."

Earlier, on the morning of May 31st, Nvidia executives announced to reporters the new Vera Rubin AI computing hardware, currently in mass production and scheduled to begin shipping to customers in Q3 2026. The new generation of chips includes the most powerful Rubin GPU to date, the Vera CPU, and a custom-designed integrated chip system for inference, designed by Groq – the company Nvidia acquired last year for $20 billion. In addition, Nvidia also stated that it is expanding its partnership with the Chinese robotics company Unitree to develop a platform for humanoid robots.

( According to vnexpress.net )

Source: https://baodongthap.vn/nvidia-ra-mat-sieu-chip-ai-cho-pc-a241589.html