This articlecontainspromotional content. Please helpimprove it by removingpromotional language and inappropriateexternal links, and by adding encyclopedic text written from aneutral point of view. See ouradvice if the article is about you and read ourscam warning in case someone asks for money to edit this article.(January 2024) (Learn how and when to remove this message)

DGX

A rack containing five DGX-1 supercomputers
Manufacturer	Nvidia
Released	2016 (2016)

TheNvidia DGX (Deep GPU Xceleration) is a series ofservers andworkstations designed byNvidia, primarily geared towards enhancingdeep learning applications through the use ofgeneral-purpose computing on graphics processing units (GPGPU). These systems typically come in a rackmount format featuring high-performancex86 server CPUs on the motherboard.

The core feature of a DGX system is its inclusion of 4 to 8Nvidia Tesla GPU modules, which are housed on an independent system board. These GPUs can be connected either via a version of theSXM socket or aPCIe x16 slot, facilitating flexible integration within the system architecture. To manage the substantial thermal output, DGX units are equipped with heatsinks and fans designed to maintain optimal operating temperatures.

This framework makes DGX units suitable for computational tasks associated with artificial intelligence and machine learning models.^{[according to whom?]}

DGX-1 servers feature 8GPUs based on thePascal orVoltadaughter cards^[1] with 128 GB of totalHBM2 memory, connected by anNVLinkmesh network.^[2] The DGX-1 was announced on 6 April 2016.^[3] All models are based on a dual socket configuration of Intel Xeon E5 CPUs, and are equipped with the following features.

• 512 GB ofDDR4-2133
• Dual 10 Gb networking
• 4 x 1.92 TBSSDs
• 3200W of combined power supply capability
• 3U Rackmount Chassis

The product line is intended to bridge the gap between GPUs andAI accelerators using specific features for deep learning workloads.^[4] The initial Pascal-based DGX-1 delivered 170teraflops ofhalf precision processing,^[5] while the Volta-based upgrade increased this to 960teraflops.^[6]

The DGX-1 was first available in only the Pascal-based configuration, with the first generation SXM socket. The later revision of the DGX-1 offered support for first generation Volta cards via the SXM-2 socket. Nvidia offered upgrade kits that allowed users with a Pascal-based DGX-1 to upgrade to a Volta-based DGX-1.^[7][8]

• The Pascal-based DGX-1 has two variants, one with a 16 coreIntel Xeon E5-2698 V3, and one with a 20 core E5-2698 V4. Pricing for the variant equipped with an E5-2698 V4 is unavailable, the Pascal-based DGX-1 with an E5-2698 V3 was priced at launch at $129,000^[9]
• The Volta-based DGX-1 is equipped with an E5-2698 V4 and was priced at launch at $149,000.^[9]

Designed as aturnkey deskside AI supercomputer, the DGX Station is atower computer that can function completely independently without typical datacenter infrastructure such as cooling, redundant power, or19 inch racks.

The DGX station was first available with the following specifications.^[10]

• Four Volta-based Tesla V100 accelerators, each with 16 GB ofHBM2 memory
• 480 TFLOPS FP16
• Single Intel Xeon E5-2698 v4^[11]
• 256 GB DDR4
• 4x 1.92 TB SSDs
• Dual 10 Gb Ethernet

The DGX station iswater-cooled to better manage the heat of almost 1500W of total system components, this allows it to keep a noise range under 35 dB under load.^[12] This, among other features, made this system a compelling purchase for customers without the infrastructure to runrackmount DGX systems, which can be loud, output a lot of heat, and take up a large area. This was Nvidia's first venture into bringinghigh performance computing deskside, which has since remained a prominent marketing strategy for Nvidia.^[13]

The Nvidia DGX-2, the successor to the DGX-1, uses sixteen Volta-based V100 32 GB (second generation) cards in a single unit. It was announced on 27 March 2018.^[14] The DGX-2 delivers 2 Petaflops with 512 GB of shared memory for tackling massive datasets and uses NVSwitch for high-bandwidth internal communication. DGX-2 has a total of 512 GB ofHBM2 memory, a total of 1.5 TB ofDDR4. Also present are eight 100 Gbit/sInfiniBand cards and 30.72 TB of SSD storage,^[15] all enclosed within a massive 10U rackmount chassis and drawing up to 10 kW under maximum load.^[16] The initial price for the DGX-2 was $399,000.^[17]

The DGX-2 differs from other DGX models in that it contains two separate GPU daughterboards, each with eight GPUs. These boards are connected by an NVSwitch system that allows for full bandwidth communication across all GPUs in the system, without additional latency between boards.^[16]

A higher performance variant of the DGX-2, the DGX-2H, was offered as well. The DGX-2H replaced the DGX-2's dual Intel Xeon Platinum 8168's with upgraded dual Intel Xeon Platinum 8174's. This upgrade does not increase core count per system, as both CPUs are 24 cores, nor does it enable any new functions of the system, but it does increase the base frequency of the CPUs from 2.7 GHz to 3.1 GHz.^[18][19][20]

Announced and released on May 14, 2020. The DGX A100 was the 3rd generation of DGX server, including 8Ampere-based A100 accelerators.^[21] Also included is 15 TB ofPCIe gen 4NVMe storage,^[22] 1 TB of RAM, and eightMellanox-powered 200 GB/s HDR InfiniBand ConnectX-6NICs. The DGX A100 is in a much smaller enclosure than its predecessor, the DGX-2, taking up only 6 Rack units.^[23]

The DGX A100 also moved to a 64 coreAMD EPYC 7742 CPU, the first DGX server to not be built with an Intel Xeon CPU. The initial price for the DGX A100 Server was $199,000.^[21]

As the successor to the original DGX Station, the DGX Station A100, aims to fill the same niche as the DGX station in being a quiet, efficient, turnkeycluster-in-a-box solution that can be purchased, leased, or rented by smaller companies or individuals who want to utilize machine learning. It follows many of the design choices of the original DGX station, such as the tower orientation, single socket CPUmainboard, a new refrigerant-based cooling system, and a reduced number of accelerators compared to the corresponding rackmount DGX A100 of the same generation.^[13] The price for the DGX Station A100 320G is $149,000 and $99,000 for the 160G model, Nvidia also offers Station rental at ~US$9000 per month through partners in the US (rentacomputer.com) and Europe (iRent IT Systems) to help reduce the costs of implementing these systems at a small scale.^[24][25]

The DGX Station A100 comes with two different configurations of the built in A100.

• Four Ampere-based A100 accelerators, configured with 40 GB (HBM) or 80 GB (HBM2e) memory,
  thus giving a total of 160 GB or 320 GB resulting either in DGX Station A100 variants 160G or 320G.
• 2.5 PFLOPS FP16
• Single 64 CoreAMD EPYC 7742
• 512 GBDDR4
• 1 x 1.92 TBNVMe OS drive
• 1 x 7.68 TB U.2 NVMe Drive
• Dual port 10 Gb Ethernet
• Single port 1 Gb BMC port

Announced March 22, 2022^[26] and planned for release in Q3 2022,^[27] The DGX H100 is the 4th generation of DGX servers, built with 8Hopper-based H100 accelerators, for a total of 32 PFLOPs of FP8 AI compute and 640 GB of HBM3 Memory, an upgrade over the DGX A100s 640GB HBM2 memory. This upgrade also increasesVRAM bandwidth to 3 TB/s.^[28] The DGX H100 increases therackmount size to 8U to accommodate the 700W TDP of each H100 SXM card. The DGX H100 also has two 1.92 TB SSDs forOperating System storage, and 30.72 TB ofSolid state storage for application data.

One more notable addition is the presence of two NvidiaBluefield 3DPUs,^[29] and the upgrade to 400 Gbit/s InfiniBand viaMellanox ConnectX-7NICs, double the bandwidth of the DGX A100. The DGX H100 uses new 'Cedar Fever' cards, each with four ConnectX-7 400 GB/s controllers, and two cards per system. This gives the DGX H100 3.2 Tbit/s of fabric bandwidth across Infiniband.^[30]

The DGX H100 has two Xeon Platinum 8480C Scalable CPUs (CodenamedSapphire Rapids)^[31] and 2 Terabytes ofSystem Memory.^[32]

The DGX H100 was priced at £379,000 or ~US$482,000 at release.^[33]

Announced May 2023, the DGX GH200 connects 32 Nvidia Hopper Superchips into a singular superchip, that consists totally of 256 H100 GPUs, 32 Grace Neoverse V2 72-core CPUs, 32 OSFT single-port ConnectX-7 VPI of with 400 Gbit/s InfiniBand and 16 dual-portBlueField-3 VPI with 200 Gbit/s ofMellanox[1][2] . Nvidia DGX GH200 is designed to handle terabyte-class models for massive recommender systems, generative AI, and graph analytics, offering 19.5 TB of shared memory with linear scalability for giant AI models.^[34]

Announced May 2023, the DGX Helios supercomputer features 4 DGX GH200 systems. Each is interconnected with Nvidia Quantum-2 InfiniBand networking to supercharge data throughput for training large AI models. Helios includes 1,024 H100 GPUs.

Announced March 2024,^[35] GB200 NVL72 connects 36 Grace Arm Neoverse V2 72-core CPUs and 72 B200 GPUs in a rack-scale design.^[36] The GB200 NVL72 is a liquid-cooled, rack-scale solution that boasts a 72-GPU NVLink domain that acts as a single massive GPU.^[37] Nvidia DGX GB200 offers 13.5 TB HBM3e of shared memory with linear scalability for giant AI models, less than its predecessor DGX GH200.

The DGX Superpod is a high performance turnkeysupercomputer system provided by Nvidia using DGX hardware.^[38] It combines DGX compute nodes with fast storage and high bandwidthnetworking to provide a solution to high demand machine learning workloads. TheSelene supercomputer, at theArgonne National Laboratory, is one example of a DGX SuperPod-based system.

Selene, built from 280 DGX A100 nodes, ranked 5th on theTOP500 list for most powerful supercomputers at the time of its completion in June 2020,^[39] and has continued to remain high in performance^{[citation needed]}. The new Hopper-based SuperPod can scale to 32 DGX H100 nodes, for a total of 256 H100 GPUs and 64 x86 CPUs. This gives the complete SuperPod 20 TB of HBM3 memory, 70.4 TB/s of bisection bandwidth, and up to 1ExaFLOP ofFP8 AI compute.^[28] These SuperPods can then be further joined to create larger supercomputers.

The Eos supercomputer, designed, built, and operated by Nvidia,^[40][41][42] was constructed of 18 H100-based SuperPods, totaling 576 DGX H100 systems, 500 Quantum-2InfiniBand switches, and 360 NVLink Switches, that allow Eos to deliver 18 EFLOPs of FP8 compute, and 9 EFLOPs of FP16 compute, making Eos the 5th fastest AI supercomputer in the world, according to TOP500 (November 2023 edition).

As Nvidia does not produce any storage devices or systems, Nvidia SuperPods rely on partners to provide high performance storage. Current storage partners for Nvidia Superpods areDell EMC,DDN,HPE,IBM,NetApp, Pavilion Data, andVAST Data.^[43]

In March 2025, Nvidia also announced the DGX Spark (previously DIGITS), a "desktop AI Supercomputer" based on Blackwell. These machines are targeted at AI researchers and programmers and have 128 GB of integrated RAM, making it possible to train or fine-tune fairly large models ("up to 200 billion parameters" with quantization). Several partner manufacturers also offer versions of the DGX Spark. It is available as of late 2025.^[44][45]

Comparison of accelerators used in DGX:^[46][47][48]

• Deep Learning Super Sampling

• Neural processing units
• GPGPU
• Nvidia products
• Parallel computing

• Articles with short description
• Short description is different from Wikidata
• Articles with a promotional tone from January 2024
• All articles with a promotional tone
• Use dmy dates from November 2023
• All articles with specifically marked weasel-worded phrases
• Articles with specifically marked weasel-worded phrases from July 2024
• All articles with unsourced statements
• Articles with unsourced statements from January 2025