When you purchase through links on our site , we may earn an affiliate commission . Here ’s how it works .
Nvidia has bring out a stock ofartificial intelligence(AI ) supercomputers that can deliver unprecedented processing world power in a portable , desktop - friendly shape .
Previously dubbed Project Digits , the powerful machines first revealedat CES 2025 in January have been rebranded as the DGX Spark and DGX Station machine . These computers are power by Nvidia ’s Blackwell Ultra chopine and promise up to a petaFLOP in processing power — upwards of 1,000 times quicker than the salutary laptop computer or high - end desktop PCs .
The DGX Spark is a little box smaller than a laptop that you could easily tuck away on a corner of your desk or fit into your bag.
Blackwell Ultra is designed for massive - scale AI training and testing , and the DGX machines predict to put that power into the hands of information scientists , AI researchers and students at a comparatively affordable price point . It ’s the equivalent of put the power of a data centre into a computer small enough to outfit on your desk .
The DGX Spark is a small box small than a laptop that you could well insert away on a corner of your desk or fit into your handbag . It stands just under 2 in ( 5 centimeter ) in peak and slightly under 6 column inch ( 15 centimeter ) in width , and at its core is the GB10 Grace Blackwell Superchip , equal to of have 1,000 AI TOPS ( trillions of operations per second ) .
concern : The 9 most powerful supercomputers in the public justly now
It also comes with 128 gigabytes of integrated system memory and Nvidia ’s full stack of AI software program , including a number of tools , libraries as well as some pretrained models . This include things like the CUDA Deep Neural connection ( cuDNN ) depository library for enhance neural connection layers during breeding and illation and a pre - trained SegFormer mannequin . The Nvidia version of the DGX Spark is uncommitted toreserve onlinestarting at $ 3,999 , although the company has articulate other model will soon be available from manufacturers like ASUS , Dell and Lenovo .
Supercomputing power in a desktop tower chassis
The DGX Station is the Spark ’s larger , more hefty sib and is tight in sizing to a professional workstation .
Built around theGB300 Grace Blackwell Ultra Desktop Superchip , it hold a staggering 748 GB of " declamatory ordered " retentiveness — memory which can be accessed by more than one processor at a sentence .
It also features Nvidia ’s ConnectX-8 SuperNIC , which enables connection connectivity at a blistering pace of up to 800 Gbit a second — fast enough to download approximately five 4 one thousand pic in a endorsement .
It also practice the NVLink - C2C Interconnect to connect internal component at 900 gigabytes per second . The DGX Station is a powerhouse design to execute large - plate AI preparation and inferencing workloads from the comfort of your background , without having to get at extra resources through the swarm .
— Supercomputer runs large and most complicated pretense of the universe ever
— World ’s fast supercomputer ' El Capitan ' goes online — it will be used to secure the US nuclear stockpile and in other classified inquiry
— Japan to begin building 1st ' zeta - year ' supercomputer in 2025 , 1,000 time more knock-down than today ’s fastest machines
The DGX Station is n’t currently useable to reserve , but Nvidia has indicate it will be available later in 2025 from pardner like Asus , Dell , HP , Lambda and Supermicro .
Jensen Huang , Nvidia ’s founder and CEO , say the raw DGX machines represent a raw next footstep in AI development .
" AI has transformed every stratum of the calculation stack . It stands to reasonableness a new class of computers would emerge , designed for AI - native developers and to run AI - native applications programme , ” he said in astatement . " With these new DGX personal AI electronic computer , AI can cross from cloud services to screen background and bound applications . "
You must confirm your public display name before commenting
Please logout and then login again , you will then be instigate to record your display name .
