During GTC 2025, Nvidia announced two new personal computing systems focused on artificial intelligence: DGX Spark and the DGX Station.
With platform technology Grace BlackwellThese devices bring to Desktop the previously restricted power to data centers, allowing data, researchers and data scientists to perform training and inferences of AI models directly on their own computers.
The novelty is useful in accessibility of accelerated computing, with support for high performance processing and optimized memory for demanding tasks. The systems will be produced by manufacturers such as ASUS, Dell, HP e Lenovo.
Accelerated Computing for AI in the desktop
O DGX Sparkpreviously called Project DIGITSis described as the smallest AI supercomputer ever developed. Its focus is on making it accessible to researchers, data scientists and students to use General and Physics Modelsallowing development and experimentation directly on a desktop.
NVIDIA seeks to expand the reach of its technology, providing more advanced innovation and tools to boost research and development in IA
Marcio Aguiar, Director of the company’s Enterprise Division for Latin America
The base of this system is the Superchip NVIDIA GB10 Grace Blackwellwhich includes a Blackwell GPU with fifth generation tensors and FP4 supportoffering up to 1,000 trillion operations per second for AI training and inference.
The connection between the GPU and the CPU occurs via NVLink-C2Cguaranteeing five times more bandwidth than the fifth generation Pie and increasing the efficiency in intensive cargo processing.
Another differential of DGX Spark is the direct integration with cloud services. The system allows locally trained models to be moved to the DGX Cloud or for other accelerated infrastructures without large code adjustments.
DGX Station: Data Center performance on desktop
O DGX Station It is a desktop AI supercomputer focused on Large -scale training and inference. The system is based on Superchip NVIDIA GB300 Grace Blackwell Ultra Desktopwhich brings 784 GB of coherent memoryoptimizing AI applications that require large processing capacity.
The main advantage of this model is the use of GPU NVIDIA Blackwell Ultra connected to CPU NVIDIA Grace through NVLink-C2Creducing latency and improving system efficiency.
In addition, DGX Station has the NVIDIA ConnectX-8 SuperNICa high -speed network plate, which allows connections to up to 800 GB/s For quick data transfers and multiple systems connectivity.
The support for NVIDIA CUDA-X™ delivers an optimized environment for developing advanced AI applications, while compatibility with the platform NVIDIA AI Enterprise enables access to optimized inference microservices.
Also read:
NVIDIA has already made available the reserves of DGX Spark on its official website, while the DGX Station will be distributed by partners as ASUS, BOXX, Dell, HP, Lambda Labs e Supermicro Still this year.
These systems are expected to boost accelerated AI adoption in various sectors, reducing data centers infrastructure dependence for certain applications.
Source: NVIDIA
Join the Adrenaline offers group
Check out the main offers of hardware, components and other electronics we find over the internet. Video card, motherboard, RAM and everything you need to set up your PC. By participating in our group, you receive daily promotions and have early access to discount coupons.
Enter the group and enjoy the promotions
Source: https://www.adrenaline.com.br/nvidia/nvidia-anuncia-supercomputadores-ia-dgx-spark-station/
