0
Enquiry €0.00
0
Basket
€0.00
Loading...
Loading...
Enquiries and orders updates!
Enquiries and orders updates! Enquiries and orders updates!
The item is out of stock.
We are sorry but the item is out of stock.

AI computers and workstations

Computing power matched to working with artificial intelligence

AI workstations deliver the computing power required for local training and deployment of LLM models, machine learning and scientific simulations. They guarantee independence from the cloud, a high level of data security and performance tailored to engineers and R&D teams.

Training and deployment of AI models

Training and deployment of AI models

Local training, fine-tuning and inference - with no cloud costs.

Data analysis and data science

Data analysis and data science

Fast work on large datasets and in ML environments.

Data under full control

Data under full control

On-premise processing, with no data sent outside.

Computers for AI

show
Sort by
Selected filters:

NVIDIA DGX Spark 1x10GbE 2x200GbE 4TB NVMe Founders Edition EU

Part number: 940-54242-0005-000

Manufacturer: NVIDIA

Processor manufacturer: Nvidia

Processor model: GB10

On-board graphics card model: NVIDIA Blackwell

Internal memory: 128 GB

Internal memory type: LPDDR5x-SDRAM

Total storage capacity: 4000 GB

Operating system installed: NVIDIA DGX OS

€5,500.00 excl. VAT
€6,765.00 incl. VAT
24h Shipment in 24H
Availability:
50+  pcs

Asus Ascent GX10 Nvidia DGX Spark Grace Blackwell 10 GB10 128GB 1TB Gen 5

Part number: 90MS0371-M00030

Manufacturer: ASUS

Processor manufacturer: Nvidia

Processor model: GB10

On-board graphics card model: NVIDIA Blackwell

Internal memory: 128 GB

Internal memory type: LPDDR5x-SDRAM

Total storage capacity: 1000 GB

Operating system installed: No

€5,000.00 excl. VAT
€6,150.00 incl. VAT
24h Shipment in 24H
Availability:
50+  pcs

Gigabyte AI TOP Atom Nvidia DGX Spark Grace Blackwell 10 GB10 128GB 4TB Gen 4

Part number: ATAGB10-9001

Manufacturer: Gigabyte

Processor manufacturer: Nvidia

Processor model: GB10

On-board graphics card model: NVIDIA Blackwell

Internal memory: 128 GB

Internal memory type: LPDDR5x-SDRAM

Total storage capacity: 4000 GB

Operating system installed: NVIDIA DGX OS

€5,200.00 excl. VAT
€6,396.00 incl. VAT
24h Shipment in 24H
Availability:
50+  pcs

Lenovo ThinkStation PGX Nvidia DGX Spark Grace Blackwell 10 GB10 128GB 4TB

Part number: 30KL0005YM

Manufacturer: Lenovo

Processor manufacturer: Nvidia

Processor model: GB10

On-board graphics card model: NVIDIA Blackwell

Internal memory: 128 GB

Internal memory type: LPDDR5x-SDRAM

Total storage capacity: 4000 GB

Operating system installed: NVIDIA DGX OS

€5,300.00 excl. VAT
€6,519.00 incl. VAT
24h Shipment in 24H
Availability:
11-20  pcs

Asus Ascent GX10 Nvidia DGX Spark Grace Blackwell 10 GB10 128GB 2TB Gen 5

Part number: 90MS0371-M000U0

Manufacturer: ASUS

Processor manufacturer: Nvidia

Processor model: GB10

On-board graphics card model: NVIDIA Blackwell

Internal memory: 128 GB

Internal memory type: LPDDR5x-SDRAM

Total storage capacity: 2000 GB

Operating system installed: No

€9,897.89 excl. VAT
€12,174.40 incl. VAT
24h Shipment in 24H
Availability:
2  pcs

MSI EdgeXpert MS-C931 Nvidia DGX Spark Grace Blackwell 10 GB10 128GB 4TB Gen 5

Part number: 9S6-C9311-32S

Manufacturer: MSI

Processor manufacturer: Nvidia

Processor model: GB10

On-board graphics card model: NVIDIA Blackwell

Internal memory: 128 GB

Internal memory type: LPDDR5x-SDRAM

Total storage capacity: 4000 GB

Operating system installed: NVIDIA DGX OS

€5,309.04 excl. VAT
€6,530.12 incl. VAT
24h Shipment in 24H
Availability:
21-30  pcs

Asus Ascent GX10 Nvidia DGX Spark Grace Blackwell 10 GB10 128GB 4TB Gen 5

Part number: 90MS0371-M000V0

Manufacturer: ASUS

Processor manufacturer: Nvidia

Processor model: GB10

On-board graphics card model: NVIDIA Blackwell

Internal memory: 128 GB

Internal memory type: LPDDR5x-SDRAM

Total storage capacity: 4000 GB

Operating system installed: No

€6,900.00 excl. VAT
€8,487.00 incl. VAT
Availability:
1  pcs

Lenovo ThinkStation PGX Nvidia DGX Spark Grace Blackwell 10 GB10 128GB 1TB

Part number: 30KL0004YM

Manufacturer: Lenovo

Processor manufacturer: Nvidia

Processor model: GB10

On-board graphics card model: NVIDIA Blackwell

Internal memory: 128 GB

Internal memory type: LPDDR5x-SDRAM

Total storage capacity: 1000 GB

Operating system installed: NVIDIA DGX OS

€4,479.95 excl. VAT
€5,510.34 incl. VAT
24h Shipment in 24H
Availability:
31-40  pcs

Gigabyte AI TOP Atom Nvidia DGX Spark Grace Blackwell 10 GB10 128GB 4TB Gen 5

Part number: ATAGB10-9000

Manufacturer: Gigabyte

Processor manufacturer: Nvidia

Processor model: GB10

On-board graphics card model: NVIDIA Blackwell

Internal memory: 128 GB

Internal memory type: LPDDR5x-SDRAM

Total storage capacity: 4000 GB

Operating system installed: NVIDIA DGX OS

€5,299.15 excl. VAT
€6,517.95 incl. VAT
24h Shipment in 24H
Availability:
5  pcs

Dell Pro Max Nvidia DGX Spark Grace Blackwell 10 GB10 128GB 4TB

Part number: N2T93

Manufacturer: Dell

Processor manufacturer: Nvidia

Processor model: GB10

On-board graphics card model: NVIDIA Blackwell

Internal memory: 128 GB

Internal memory type: LPDDR5x-SDRAM

Total storage capacity: 4000 GB

Operating system installed: NVIDIA DGX OS

€5,378.79 excl. VAT
€6,615.91 incl. VAT
24h Shipment in 24H
Availability:
21-30  pcs

Dell Pro Max Nvidia DGX Spark Grace Blackwell 10 GB10 128GB 2TB

Part number: T9WMV

Manufacturer: Dell

Processor manufacturer: Nvidia

Processor model: GB10

On-board graphics card model: NVIDIA Blackwell

Internal memory: 128 GB

Internal memory type: LPDDR5x-SDRAM

Total storage capacity: 2000 GB

Operating system installed: NVIDIA DGX OS

€5,221.18 excl. VAT
€6,422.05 incl. VAT
24h Shipment in 24H
Availability:
3  pcs

Lenovo ThinkStation PGX Nvidia DGX Spark Grace Blackwell 10 GB10 128GB 4TB

Part number: 30KL0003GF

Manufacturer: Lenovo

Processor manufacturer: Nvidia

Processor model: GB10

On-board graphics card model: NVIDIA Blackwell

Internal memory: 128 GB

Internal memory type: LPDDR5x-SDRAM

Total storage capacity: 4000 GB

Operating system installed: NVIDIA DGX OS

€6,000.00 excl. VAT
€7,380.00 incl. VAT
Availability:
11-20  pcs

Dell Pro Max Nvidia DGX Spark Grace Blackwell 10 GB10 128GB 4TB

Part number: CD2J8

Manufacturer: Dell

Processor manufacturer: Nvidia

Processor model: GB10

On-board graphics card model: NVIDIA Blackwell

Internal memory: 128 GB

Internal memory type: LPDDR5x-SDRAM

Total storage capacity: 4000 GB

Operating system installed: NVIDIA DGX OS

€5,533.13 excl. VAT
€6,805.75 incl. VAT
24h Shipment in 24H
Availability:
21-30  pcs

HP ZGX Nano G1n Nvidia DGX Spark Grace Blackwell 10 GB10 128GB 4TB

Part number: CZ9K4ET#ABD

Manufacturer: HP

Processor manufacturer: Nvidia

Processor model: GB10

On-board graphics card model: NVIDIA Blackwell

Internal memory: 128 GB

Internal memory type: LPDDR5x-SDRAM

Total storage capacity: 4000 GB

Operating system installed: Linux

€5,602.23 excl. VAT
€6,890.74 incl. VAT
Availability:
9  pcs

Dell Pro Max Nvidia DGX Spark Grace Blackwell 10 GB10 128GB 2TB

Part number: 94XGD

Manufacturer: Dell

Processor manufacturer: Nvidia

Processor model: GB10

On-board graphics card model: NVIDIA Blackwell

Internal memory: 128 GB

Internal memory type: LPDDR5x-SDRAM

Total storage capacity: 2000 GB

Operating system installed: NVIDIA DGX OS

€5,186.00 excl. VAT
€6,378.78 incl. VAT
24h Shipment in 24H
Availability:
2  pcs

Lenovo ThinkStation PGX Nvidia DGX Spark Grace Blackwell 10 GB10 128GB 1TB

Part number: 30KL0002GF

Manufacturer: Lenovo

Processor manufacturer: Nvidia

Processor model: GB10

On-board graphics card model: NVIDIA Blackwell

Internal memory: 128 GB

Internal memory type: LPDDR5x-SDRAM

Total storage capacity: 1000 GB

Operating system installed: NVIDIA DGX OS

€4,714.01 excl. VAT
€5,798.23 incl. VAT
24h Shipment in 24H
Availability:
7  pcs

HP ZGX Nano G1n Nvidia DGX Spark Grace Blackwell 10 GB10 128GB 4TB

Part number: CZ9K0ET#ABF

Manufacturer: HP

Processor manufacturer: Nvidia

Processor model: GB10

On-board graphics card model: NVIDIA Blackwell

Internal memory: 128 GB

Internal memory type: LPDDR5x-SDRAM

Total storage capacity: 4000 GB

Operating system installed: Linux

€9,307.18 excl. VAT
€11,447.83 incl. VAT
Availability:
41-50  pcs

Dell Pro Max Nvidia DGX Spark Grace Blackwell 10 GB10 128GB 4TB

Part number: W2KP4

Manufacturer: Dell

Processor manufacturer: Nvidia

Processor model: GB10

On-board graphics card model: NVIDIA Blackwell

Internal memory: 128 GB

Internal memory type: LPDDR5x-SDRAM

Total storage capacity: 4000 GB

Operating system installed: NVIDIA DGX OS

€5,351.38 excl. VAT
€6,582.20 incl. VAT
24h Shipment in 24H
Availability:
41-50  pcs

Lenovo ThinkStation PGX Nvidia DGX Spark Grace Blackwell 10 GB10 128GB 4TB

Part number: 30KL000AFC

Manufacturer: Lenovo

Processor manufacturer: Nvidia

Processor model: GB10

On-board graphics card model: NVIDIA Blackwell

Internal memory: 128 GB

Internal memory type: LPDDR5x-SDRAM

Total storage capacity: 4000 GB

Operating system installed: NVIDIA DGX OS

€7,738.75 excl. VAT
€9,518.66 incl. VAT
Availability:
21-30  pcs

Dell Pro Max Nvidia DGX Spark Grace Blackwell 10 GB10 128GB 4TB

Part number: 7MFXX

Manufacturer: Dell

Processor manufacturer: Nvidia

Processor model: GB10

On-board graphics card model: NVIDIA Blackwell

Internal memory: 128 GB

Internal memory type: LPDDR5x-SDRAM

Total storage capacity: 4000 GB

Operating system installed: NVIDIA DGX OS

€5,533.13 excl. VAT
€6,805.75 incl. VAT
24h Shipment in 24H
Availability:
21-30  pcs

Lenovo ThinkStation PGX Nvidia DGX Spark Grace Blackwell 10 GB10 128GB 1TB

Part number: 30KL000CFC

Manufacturer: Lenovo

Processor manufacturer: Nvidia

Processor model: GB10

On-board graphics card model: NVIDIA Blackwell

Internal memory: 128 GB

Internal memory type: LPDDR5x-SDRAM

Total storage capacity: 1000 GB

Operating system installed: NVIDIA DGX OS

€5,361.52 excl. VAT
€6,594.67 incl. VAT
Availability:
21-30  pcs

Acer Veriton GN100 Nvidia DGX Spark Grace Blackwell 10 GB10 128GB 4TB

Part number: DT.R6LEH.002

Manufacturer: Acer

Processor manufacturer: Nvidia

Processor model: GB10

On-board graphics card model: NVIDIA Blackwell

Internal memory: 128 GB

Internal memory type: LPDDR5x-SDRAM

Total storage capacity: 4000 GB

Operating system installed: NVIDIA DGX OS

€6,147.78 excl. VAT
€7,561.77 incl. VAT
Availability:
11-20  pcs

MSI EdgeXpert MS-C931 Nvidia DGX Spark Grace Blackwell 10 GB10 128GB 4TB Gen 5

Part number: EDGEXPERT-32SEU

Manufacturer: MSI

Processor manufacturer: Nvidia

Processor model: GB10

On-board graphics card model: NVIDIA Blackwell

Internal memory: 128 GB

Internal memory type: LPDDR5x-SDRAM

Total storage capacity: 4000 GB

Operating system installed: NVIDIA DGX OS

€5,423.96 excl. VAT
€6,671.47 incl. VAT
24h Shipment in 24H
Availability:
11-20  pcs

HP ZGX Nano G1n Nvidia DGX Spark Grace Blackwell 10 GB10 128GB 1TB

Part number: D4DR6ET#AKD

Manufacturer: HP

Processor manufacturer: Nvidia

Processor model: GB10

On-board graphics card model: NVIDIA Blackwell

Internal memory: 128 GB

Internal memory type: LPDDR5x-SDRAM

Total storage capacity: 1000 GB

Operating system installed: Linux

€7,404.62 excl. VAT
€9,107.68 incl. VAT
Availability:
11-20  pcs

AI Computers – Buying Guide

AI computers are dedicated workstations designed for local training, fine-tuning, and running artificial intelligence models. The choice of a computer for working with artificial intelligence depends on how the models will be used - from prototyping and fine-tuning large language models, to image generation, to real-time inference, which involves handling queries to a ready model with minimal latency. In this guide, you will find a comprehensive comparison of parameters, manufacturers, and configurations available from Senetic, which will help IT Managers, developers, and purchasing teams select hardware optimal for performance, scalability, and budget.

What are AI Computers

An AI computer is a workstation or compact desktop system optimised for artificial intelligence tasks: machine learning, training neural networks and models on proprietary datasets, and inference. An AI computer typically has a more complex architecture than a regular computer - key components for AI computers include graphics card, RAM, and processor with NPU.

The basic operating principles are based on three types of computing units:

  • GPU (Graphics Processing Unit) - a graphics chip responsible for parallel processing of large data matrices, including calculations on weights and gradients of artificial intelligence models. The GPU divides computations into thousands of simultaneous threads, thereby reducing the training and running time of AI models from weeks to hours. A key parameter when choosing an AI computer is the amount of GPU memory (VRAM) as it determines how large a model can fit in the card's memory without swapping data with the disk.
  • NPU (Neural Processing Unit) - a specialised processor for inference operations. Dedicated NPU processors offload the system in AI tasks, providing low latency and high efficiency in local data processing. The presence of NPU units allows for efficient local data processing in AI.
  • SoC (System on Chip) - an integrated chip that combines a processor (CPU), graphics processor (GPU), and unified memory in one chip. This allows all components to use a shared pool of memory instead of exchanging data between separate chips. This eliminates the transfer bottleneck that occurs in traditional computers at the junction of CPU, GPU, and separate RAM/VRAM. This approach is used by compact AI computers, enabling work with large models while consuming less power and occupying less space than a traditional workstation.

The benefits of using AI computers in a business environment include:

  • full control over data,
  • lower latency than cloud solutions,
  • no dependence on internet bandwidth,
  • lower operational costs in the long term,
  • enabling local model deployment - for many companies, this means new opportunities, from prototyping to production deployment, without the need for cloud processing.

The difference between local and cloud processing comes down to a trade-off: the cloud offers scalability but incurs ongoing costs and requires data transfer. Local AI computers eliminate these limitations - especially with frequent and intensive model training or inference.

AI Computer vs Regular Computer with Graphics Card

A regular desktop computer with a powerful graphics card and a professional workstation are two different classes of devices. The graphics card is the most important component of an AI computer, but it alone does not define an AI computer.

Memory Architecture

The memory architecture directly affects how large models can be handled without losing performance. In SoC-based computers with unified memory, the CPU and GPU share a common pool of memory, allowing for work with large language models without splitting them into smaller segments and without losing performance when switching data between chips. This is a real advantage for tasks requiring large amounts of memory, e.g., working with models in the range of tens of billions of parameters, which in standard architecture would require several interconnected graphics cards.

Performance

Performance in AI tasks varies dramatically depending on the class of hardware. Consumer-grade graphics cards achieve AI performance measured in thousands of TOPS (Tera Operations Per Second), which is sufficient for many inference tasks and fine-tuning smaller models. Dedicated AI computers, based on architectures optimised for computation, offer computational power unavailable to consumer-grade graphics cards.

Process Acceleration

AI computers accelerate both processes related to artificial intelligence and creative tasks from rendering graphics, through video processing, to image generation in AI model-based tools. However, for professional model training, a dedicated workstation using the architecture and unified memory described above is sufficient. Consumer-grade graphics cards, even the most powerful, do not have enough memory or bandwidth for this scale of tasks.

Cooling

AI computers often have higher power and cooling requirements due to prolonged loads, which is another element that distinguishes them from standard PCs.

Types of AI Computers by Application

The choice of an AI computer should consider the type of tasks performed. Different stages of working with AI models - from training from scratch, through fine-tuning, to production deployment - require different hardware configurations.

Computers for Training Models

Training large neural networks from scratch is one of the most computationally demanding AI applications. The performance of a workstation primarily depends on the power of the GPU in the computations used during training, the capacity and bandwidth of the GPU memory, and the speed of data delivery. Depending on the size of the model, GPUs equipped with several dozen gigabytes of VRAM or systems with large unified memory may be useful. Large training datasets also justify the use of fast NVMe disks with several terabytes of capacity and a large amount of RAM. PCIe Gen4 or Gen5 can further limit communication bottlenecks, especially in configurations with multiple GPUs. However, specific requirements depend on the type and scale of the model being trained.

Computers for Fine-tuning Models

Fine-tuning existing AI models requires a balanced configuration. A GPU or a Blackwell-type chip with a medium/high level of VRAM, a fast NVMe SSD for reading and writing checkpoints, and a large capacity of RAM form an optimal base. The speed of writing to disk and memory bandwidth directly affects the time of each training cycle.

Computers for AI Inference

Running ready AI models requires optimisation for response speed (latency) and energy efficiency. Both GPU cards and NPU processors perform well here. RTX performance is crucial for local AI image generation, and compact SoC systems with unified memory achieve excellent results with low energy consumption.

Key Parameters of AI Computers

Technical parameters determine what AI tasks a computer will be able to perform efficiently. Below is a discussion of key specifications to pay attention to when making a selection.

Disk Capacity - 1 TB, 2 TB or 4 TB

NVMe disks are crucial for fast data loading in AI applications. SSDs with a capacity of 512 GB are recommended for entry-level AI computers - 512 GB NVMe SSDs are versatile for AI in simpler inference tasks.

For professional applications:

  • 1 TB - sufficient for inference and prototyping with single models
  • 2 TB - optimal for fine-tuning when data and checkpoints are stored locally
  • 4 TB - recommended when training models and working with large datasets

Connectivity 10 GbE and 200 GbE

Fast network connectivity is very important in AI environments, especially when several computers collaborate within a single cluster or use data stored on local servers. Efficient workstations can offer 10 GbE ports, and more advanced configurations may also include interfaces with bandwidth reaching 200 Gb/s.

Such fast connections facilitate the exchange of large amounts of data between systems, synchronisation of computations, and distribution of tasks among several units. This is particularly important when working with very large models, which due to memory or computational power requirements may necessitate the use of more than one computer.

Wired connectivity can be complemented by Wi-Fi 7 and Bluetooth 5.4.

RAM and VRAM

In a conventional PC it is the graphics card's VRAM — typically 16–32 GB — that determines how large a model you can run, while system RAM (16 GB for everyday work, 32 GB or more for demanding projects) handles everything else. AI computers built on NVIDIA GB10 work differently: 128 GB of unified memory forms a single pool shared by the CPU and GPU, so both the VRAM ceiling and the copying of data between memories disappear.

Processors and Graphics Cards

Modern AI computers rely on two architectural approaches:

  • Traditional platforms with a multi-core general-purpose processor and a dedicated graphics card
  • SoC platforms with a multi-core processor integrated with a graphics chip on one chip

AI computers should have dedicated NPU (Neural Processing Unit) units regardless of the platform; the presence of NPU ensures efficient offloading of the processor and graphics card during routine AI operations, such as speech recognition, image processing, or the operation of AI assistants in the operating system. This allows the computer to remain efficient even with background tasks without burdening the main computing units.

AI Computer Manufacturers Available at Senetic

Senetic offers AI computers from leading manufacturers such as NVIDIA, Lenovo, HP, Dell, Asus, MSI, Gigabyte, and Acer.

NVIDIA

NVIDIA offers its own AI computer, designed as a complete solution for local artificial intelligence computations. The device combines computational performance with a ready working environment. The operating system includes a full software stack for working with AI models, allowing you to start working without additional configuration. This is a proposal for teams that want to launch AI projects immediately after unpacking the hardware, without time-consuming environment preparation.

Lenovo

Lenovo offers AI computers as part of its line of workstations designed for business applications. The models feature extensive data security functions, including disk encryption and a trusted platform module, which is important when working with sensitive data or compliance requirements. Variants with different disk capacities are available, tailored to the scale of AI projects being conducted.

HP

HP introduces AI computers as an extension of its line of workstations for business. The designs are compact, intended for deployments in office and laboratory spaces, where both performance and limited footprint matter. The manufacturer emphasises enterprise-class reliability and cooling optimisation in a small form factor.

Dell

Dell offers AI computers as part of a series designed for companies with varying scales of deployment. A significant element of the offering is the device fleet management ecosystem, which facilitates administration of a larger number of workstations within an organisation. The manufacturer is developing this line as part of a broader strategy for computers supported by hardware AI accelerators.

ASUS

ASUS offers AI computers designed for long-term, continuous computational loads. The design emphasises cooling solutions tailored for working with large datasets and AI models over extended periods. This is a proposal aimed at teams conducting regular, repetitive computational tasks.

MSI

MSI specialises in AI computers intended for edge computing applications, meaning data processing close to where it is generated, rather than sending it to a central server or cloud. The manufacturer's offering is particularly aimed at industrial applications, where real-time inference is required directly on-site. The compact format of the devices facilitates deployments in production, warehouse, and IoT environments.

Gigabyte

Gigabyte offers AI computers complemented by its own software tools supporting performance optimisation during fine-tuning and inference of models. Variants with different SSD disk capacities are available, tailored to various project scales. The devices can be interconnected, allowing for increased resources when working with larger models.

Acer

Acer targets its offering of AI computers primarily at small and medium-sized enterprises beginning to implement artificial intelligence in their processes. The emphasis is on affordability while maintaining a compact, desktop design. This solution is suitable for both local inference and early stages of model prototyping.

FAQ - Frequently Asked Questions

Is a regular computer with a powerful graphics card sufficient for working with AI?

It depends on the scale of the task. For inference on smaller models, image generation, video editing with AI acceleration, or daily work with code assistants, yes, a computer with a powerful graphics card and at least 16 GB of RAM is sufficient. However, for training your own models, working with large language models over 70 billion parameters, or production inference for an entire team, you need a dedicated AI workstation with unified memory and SoC architecture.

How large a language model can fit in 128 GB of unified memory?

Models with 70-80 billion parameters typically fit in ~30-40 GB of GPU memory when using memory compression. In 128 GB of unified memory, you can comfortably run models up to ~200 billion parameters. Models with 400 billion parameters require connecting two devices via high-bandwidth network ports, resulting in a total of 256 GB of unified memory.

Do AI computers from different manufacturers differ in performance?

Models based on the same SoC (e.g., NVIDIA DGX Spark, Lenovo ThinkStation PGX, Gigabyte AI TOP Atom, MSI EdgeXpert MS-C931) offer similar computational performance and the same amount of unified memory. Differences relate to cooling quality, network interfaces, security features, pre-installed software, warranty conditions, and availability of technical support.

What operating system does NVIDIA DGX Spark support?

NVIDIA DGX Spark runs on a Linux-based system, with a pre-installed full software stack for working with AI (computing environments, machine learning libraries, analytical notebooks). Windows is not supported on this hardware platform.

Can two AI computers be connected into one cluster?

Some models in this category, including Lenovo ThinkStation PGX, Gigabyte AI TOP Atom, NVIDIA DGX Spark, or MSI EdgeXpert MS-C931, are equipped with high-bandwidth network ports that allow two units to be connected into a cluster.

How much power does an AI computer consume?

The typical power consumption for compact AI stations based on SoC architecture is around 240 W, comparable to a powerful desktop computer, but with significantly higher AI performance per watt of energy consumed. In configurations with a traditional graphics card, consumption may be higher. The high-end card itself consumes up to about 450 W, and the entire system 600-800 W under full load.

When does local AI processing have an advantage over the cloud?

Local AI stations outperform the cloud when you need: low latency (real-time inference), full control over data and compliance with regulations, independence from internet bandwidth, and when you are intensively and regularly performing fine-tuning or inference of models. In the long term, the operational cost of a local station can be lower than the fees for cloud computing resources with continuous use. Many companies are exploring a hybrid model, local processing on a daily basis, and cloud for scaling peak loads.

What disk capacity should I choose for working with AI models?

1 TB is sufficient for inference and prototyping with a few models. 2-4 TB is recommended for fine-tuning and training. A 512 GB disk is only suitable for the simplest applications. The priority should be the speed of the disk; a fast NVMe drive is more important than capacity itself, as it reduces data loading times and checkpoint writing.

Stay in touch!
Join our newsletter and stay up to date with our promotions and news.

By clicking "Subscribe" you agree to receive marketing information at your e-mail address. Your personal data will be administered by Senetic SA with its registered office in ul. Przewozowa 32, 44-100 Gliwice, Poland. You have the right to demand access to personal data and to rectify, erase or limit the processing, as well as to object to the processing. Detailed information on the processing of personal data can be found in the privacy policy.

Incorrect e-mail address
This e-mail address has already been confirmed.
This e-mail address has not been confirmed yet.
Thank you!
Check your inbox - you will receive a message from us asking you to confirm your subscription to the newsletter.
Reseller portal
Everything you are looking for
Microsoft
Full range of Microsoft solutions
Please wait ...
Compare (0/4)