NVIDIA has announced the launch of DGX Spark, paving the way for what it calls “personal AI supercomputers”a new group of machines for artificial intelligence development and research from the computing desktop.
If DGX Spark is the general design proposed by NVIDIA, relevant partners of the graphics giant have also announced the availability of custom machines on this platform. This includes models from ASUS, GIGABYTE or Dell.
DGX Spark, the computer of the AI era
The development of AI is not limited to gigantic data centers and the computing desktop also has its relevance to advance the revolution. That is NVIDIA’s goal: put an AI server on the workbench available to researchers.
The machines on this platform have a surprisingly compact size, which could pass for a typical mini-PC with dimensions of only 150 X 150 x 50.5 mm. Its internal engine is the new Grace GB10 chip created by NVIDIA in collaboration with MediaTek and includes a Blackwell GPU connected to a 20-core Grace CPU.
Inside the chassis, the chips are connected to a 128 Gbyte memory array and accommodate up to 4 TB of storage with solid state drives. It has support for Wi-Fi 7 and Bluetooth 5.3 and a set of ports that includes four USB4 ports (40 Gbps), 1 10 GbE LAN (ConnectX-7 smart NIC) and an HDMI 2.1 for connection to external displays. Total system power consumption is 170 watts.
This solution has a power of 1 petaflop (up to 1000 trillion operations per second of AI computing) to prototype, tune and run AI models. NVIDIA claims that a single unit can run models with a size of up to 200 billion parameters.
DGX Spark runs on NVIDIA’s Linux-based DGX operating system and can operate standalone or connected to a host PC running Windows or macOS. Its reference Price is $3,000.
Customized machines offer
Several manufacturers support the launch of this platform with customized machines, which will also be available from this week. Among them we can highlight the ASUS Ascent GX10, the EdgeXpert Desktop AI from MSI, the AI TOP ATOM from GIGABYTE or the Dell Pro Max.
They all use a similar design, tremendously compact in size, but with internal power thanks to the NVIDIA accelerator. They also share the same objectives: to put the capabilities of a petaFLOP-scale AI supercomputer right on desktops from developers, AI researchers and data scientists around the world.
