**NVIDIA announces the new DGX Spark product, positioned as a "supercomputer for the masses" for working with intelligent systems**
NVIDIA company officially announced the start of sales and distribution of its new DGX Spark system, which will be available from November 18, 2024. The system is positioned as a powerful instrument for training and launching neural networks, oriented as for developers, so for researchers and on students.
**Price and configurations** The DGX Spark price varies depending on the configuration. The basic version, with the Ampere chip with 48 cores and 128 GB of GPU memory, costs $3999, and the base one — with an A100 GPU and 32 GB of GPU memory — in $1999. The product competes with the solutions of such companies as Google (Coral) and Amazon (Trainium), but, as experts note, NVIDIA has a key advantage in the CUDA ecosystem — a software platform for parallel computing.
**Technical characteristics** DGX Spark is based on an AMD Ryzen 7 processor, Ampere/A100 GPU, LPDDR5 memory for 32 or 128 GB and NVMe drives with 1 TB. The system supports the launch of various neural network models, including popular architectures such as Stable Diffusion and Mistral. The main advantage is that developers can test their applications locally, minimizing the load on server services.
**Software and support** NVIDIA provides educational programs and support for DGX Spark users. Specialized courses, developed in cooperation with Stanford HAI, as well as technical assistance in corporate programs are available.
**Expert reaction and users** Jim Keller, analyst Gartner, noted that Ampere gives customers an advantage in the CUDA ecosystem, which accelerates integration. One limitation in the number of PCIe slots and a price policy — $11 per gigabyte of GPU memory — may become a barrier for mass adoption.
**Quotes from key persons** ̄енгл из NVIDIA заявил: *"We strive to make the power of AI accessible to everyone. DGX Spark is another step in the democratization of computing"*.
Ilon Mask said: *"Such systems open the door to a new era of AI-oriented solutions"*.
**Perspectives of use** According to experts, DGX Spark will find application in small and medium-sized businesses, and also in scientific institutions. One of its key advantages is the versatility of the NVIDIA ecosystem and the cost of GPU memory.
