NVIDIA has revolutionized the world of artificial intelligence With the launch of its new high-performance personal computers, the DGX Spark and DGX StationThese devices are designed for the development and deployment of AI models without relying on large data center infrastructures, bringing this technology to a more accessible and compact form factor.
Thanks to architecture Grace Blackwell, which until now was only available in data centers, NVIDIA has managed to integrate outstanding performance into desktop systems. In this article, we'll explore all the features, specifications, and benefits offered by these new AI workstations.
DGX Spark: The most compact AI supercomputer
El DGX Spark, formerly known as Project DIGITS, is the world's smallest AI supercomputer. This mini desktop computer is designed for researchers, data scientists, and robotics developers looking for a powerful device without the need for a data center infrastructure.
It is equipped with the NVIDIA GB10 Grace Blackwell superchip, which includes a GPU with fifth-generation Tensor cores and FP4 precision. Thanks to this technology, it can run up to 1.000 trillion operations per second (TOPS), making it an ideal solution for developing advanced language models such as those that can be optimized with technologies such as Raspberry Pi.
DGX Spark Technical Specifications
- Processor: NVIDIA GB10 Grace Blackwell
- Memory: 128GB of LPDDR5X unified memory
- Storage: Up to 4 TB NVMe SSD
- Interconnection: NVLink-C2C technology that increases bandwidth fivefold compared to PCIe 5.0
- Performance: 1.000 trillion operations per second of AI computing
El DGX Spark It is compact and easy to integrate into any workspace. Its main advantage is the ability to develop and optimize AI models locally, with the option to seamlessly migrate to NVIDIA DGX Cloud or any other NVIDIA-accelerated infrastructure. Developers can create innovative solutions such as a smartlamp with Arduino using this advanced technology.
DGX Station: Extreme Power on a Desktop
While the DGX Spark is a compact option, NVIDIA has also introduced the DGX Station, a much more powerful workstation designed for more demanding workloads. This computer is based on the NVIDIA GB300 Grace Blackwell Ultra superchip and is focused on training and inference of large-scale AI models, similar to 3D printers, which require large processing capabilities.
Among its most notable features we find:
- Superchip GB300 with Blackwell Ultra architecture
- Memory: 784GB unified memory
- Advanced Connectivity: ConnectX-8 SuperNIC smart network card with support for up to 800 Gb/s
- Performance: 20 petaflops in AI
The DGX Station is a must-have tool for researchers looking to take their AI projects to the next level, just like the 3D printers in the field of digital manufacturing.
Availability and prices
NVIDIA has confirmed that the DGX Spark is now available for reservation through its website with an initial price of $2.999 in its basic configuration, which includes a 20-core processor and 128 GB of memory.
In turn, the DGX Station will be available by the end of the year and will be distributed by manufacturers such as ASUS, Dell, HP, and Supermicro. Pricing has not yet been revealed, but it is expected to be considerably higher than the DGX Spark due to its increased capacity and performance. Those interested in taking full advantage of these technologies may want to consider options such as official printer distributors to equip their laboratories.
The arrival of NVIDIA DGX Spark and DGX Station It marks a milestone in the democratization of artificial intelligence, allowing more professionals, researchers, and enthusiasts to access high-performance technologies without the need for massive infrastructure. Its combination of powerful hardware and optimized software opens up new possibilities for the development and deployment of AI models in local environments.
