AI, rendering and computing
High-performance AI servers for data centres
Data centre GPU servers for training and running the largest AI models. We offer the NVIDIA HGX B300 platform with eight Blackwell Ultra GPUs and 2,160 GB of HBM3e memory in the GIGABYTE G894-SD3 rack server, the ASUS ExpertCenter Pro ET900N compute station with NVIDIA GB300 and rack servers with four RTX 5090 or two A100 80 GB cards for inference and model fine-tuning.
Which models fit: an HGX B300 server with 2,160 GB of HBM3e serves models with hundreds of billions up to roughly a trillion parameters and is suitable for training and full fine-tuning of large models. Servers with 128–160 GB of VRAM handle inference of models of up to roughly 120 billion parameters for dozens of users. Several servers can be connected into an AI cluster.
Who it is for: data centres, AI service providers, research teams and companies that train their own models. We supply HGX and GB300 servers to order – we prepare the configuration, price and delivery date after a consultation. You can find more about the process on the AI servers page and in the article AI server for business, work and home.
- 8 systemsfrom 442 290 Kč incl. VAT
- Pre-installedswitch it on and the server is ready to work
- Made to orderwe adjust the configuration to your needs
Server comparison
Prices excl. VAT, with VAT below. We will adjust the configuration to suit you.
| Server | GPU | VRAM | Processor | RAM | Power draw | Price excl. VAT |
|---|---|---|---|---|---|---|
| AI server PC Praha Rome Dual | 8 GPUs of your choice | 128–768 GB | 2× AMD EPYC 7643 | 512–1024 GB | — | from 731 893 Kč885 590 Kč incl. VATmade to order |
| AI server 4× RTX 5090 | 4× RTX 5090 | 128 GB | AMD EPYC 7643 | 512 GB DDR4 ECC | ≈ 2.7 kW | 990 909 Kč1 199 000 Kč incl. VATmade to order |
| AI server HPE ProLiant DL385 Gen10 | 2× A100 80 GB | 160 GB | 2× AMD EPYC 7502 | 512 GB DDR4 ECC | ≈ 1.2 kW | 1 020 510 Kč1 234 817 Kč incl. VATmade to order |
| AI server PC Praha Turin Dual | 4–16 GPUs of your choice | 64–1536 GB | 2× AMD EPYC 9355 | 768–1536 GB | — | from 1 443 793 Kč1 746 990 Kč incl. VATmade to order |
| ASUS ExpertCenter Pro ET900N G3 | 1× GB300 Grace Blackwell Ultra | 252 GB | NVIDIA Grace (72 jader ARM Neoverse V2) | 748 GB coherent (HBM3e + LPDDR5X) | — | 2 491 727 Kč3 014 990 Kč incl. VATmade to order |
| GIGABYTE G894-SD3 | 8× B300 (HGX) | 2160 GB | 2× Intel Xeon 6776P | 1024 GB DDR5 | — | 12 865 694 Kč15 567 490 Kč incl. VATmade to order |
| GIGABYTE G894-SD3 | 8× B300 (HGX) | 2160 GB | 2× Intel Xeon 6776P | 2048 GB DDR5 | — | 13 920 570 Kč16 843 890 Kč incl. VATmade to order |
VRAM = total graphics memory of all cards. Power draw at full load is approximate.
What the servers are for
- AI and LLMs
Language models in your company
An internal chatbot, assistant or working with company documents (RAG). Your data never leaves the company and you pay no monthly fees for cloud GPUs.
- Images, video, 3D
Generation and rendering
Image and video generation (FLUX.1, SDXL, ComfyUI), rendering in Blender and workloads that need lots of cores and memory.
- Bonus
Mining in idle time
When the server isn't running your workloads, it can mine cryptocurrencies on its graphics cards. See how much the cards earn on our graphics card profitability page.
AI server FAQ
What is the difference between an NVIDIA HGX B300 server and a server with RTX cards?
The HGX B300 has eight Blackwell Ultra GPUs linked by a fast NVLink interconnect and 2,160 GB of HBM3e memory – it is built for training and running the largest models. Servers with RTX 5090 or A100 are more affordable and are suited mainly to inference and fine-tuning of models up to roughly 120 billion parameters.
What does a GPU server need in a data centre?
A rack of the right height, a strong power supply, sufficient cooling and a fast network connection. We check the requirements of the specific configuration in advance. If you don't have your own data centre, we can place the server in our hosting.
Can the servers be linked into an AI cluster?
Yes. HGX servers are commonly linked into clusters over a high-speed network. We design the topology, network components and storage according to the size of the project.
How long does delivery of an HGX B300 or GB300 server take?
We supply these servers to order. After a consultation we prepare the configuration, price and delivery date based on current availability from the manufacturer.
Are these servers suitable for training your own models?
Yes. The HGX B300 with 2,160 GB of HBM3e is designed for training and full fine-tuning of large models; the GB300 with 748 GB of coherent memory for development and fine-tuning of large models by a single team. For smaller fine-tuning (LoRA), servers with RTX 5090 or A100 are enough.
Discuss a configuration
Tell us which models or programs you want to run, how many users and what your budget is. We will propose a configuration and send you a quote.
- Phone+420 792 480 835
- Emailobchod@pcpraha.cz
- Shop and service centreRostovská 260/2b, Praha 10 · Mon–Fri 9–17