PC Praha Turin Dual AI server with eight graphics cards and two EPYC processors

AI server PC Praha Turin Dual – 2× EPYC 9005, up to 16 GPUs

4× RTX 5080 16 GB (64 GB VRAM) / 768 GB DDR5 ECC / 4 TB NVMe
1 786 590 Kč
Skip to product information
PC Praha Turin Dual AI server with eight graphics cards and two EPYC processors
  • PC Praha Turin Dual AI server with eight graphics cards and two EPYC processors
  • PCIe riser cabling to the graphics cards
  • Fitting the graphics cards – front view
  • AI server packed for shipping
AI server PC Praha Turin Dual – 2× EPYC 9005, up to 16 GPUs

PC Praha · AI and GPU servers

AI server PC Praha Turin Dual – 2× EPYC 9005, up to 16 GPUs

Specifications by configuration

Graphics cards
4–16 GPUs
Graphics memory
64–1536 GB
Processor
2× AMD EPYC 9355
Memory
768–1536 GB
Form factor
7U rack for 8 GPUs
1 476 521 Kč excl. VAT
1 786 590 Kč incl. 21% VAT
Graphics cards
RAM
Storage

Assembly and testing: usually 5–7 working days. We often have the server in stock and ship it straight away – we always confirm the date individually.

Graphics cards
RAM
Storage
  • 24-month warranty, claims handled directly by us
  • We help with configuration and software
  • We can run it in our data centre on request

What the build includes

Graphics cards
4× RTX 5080 16 GB 64 GB VRAM in total
Processor
2× EPYC 9355 · Turin 32C
Motherboard
ASRock TURIN2D24XGM
Memory
12× 64 GB DDR5 ECC RDIMM 768 GB in total
Storage
4 TB NVMe
Power supply
4× 2000 W power supply, redundant
Chassis
7U rack for 8 GPUs
GPU risers
4× PCIe riser
Cooling
CPU cooler and server fans
Assembly
Assembly, OS installation and 72-hour stress test

The price above is for the complete server including assembly. We are happy to quote a different CPU, network card or chassis.

Our flagship platform: two AMD EPYC 9005 processors, up to 1.5 TB of DDR5 ECC memory and four to eight graphics cards – or sixteen RTX Pro 6000 cards with 1.5 TB of graphics memory.

Full description, video and specifications

Which models this build can run

The selected configuration has 64 GB of graphics memory (VRAM). It comfortably fits, for example, these open language models:

  • Llama 3.3 70B / Qwen3-72B48 GB
  • Seed-OSS-36B28 GB
  • DeepSeek-R2 32B24 GB
  • Qwen3-30B-A3B MoE20 GB
  • Gemma 3 27B18 GB
  • Mistral Small 3 24B14 GB
  • Qwen3-14B12 GB

The larger model Mistral Large 2 123B needs roughly 70 GB of VRAM.

Indicative estimate for quantised models (mostly Q4) with headroom for context. Actual memory needs depend on context length and the number of concurrent users.

What we take care of

  • Custom pre-installation

    We pre-install Linux, NVIDIA drivers, CUDA, Docker and the software for your workload (e.g. vLLM or Ollama) – you can start working as soon as you switch it on.

  • 72 hours under load

    Before handover every server runs at full load for 72 hours while we check stability, temperatures and ECC memory errors.

  • Service within 24/48 hours

    We keep key spare parts in stock in Prague. Service visits and part replacement within 24/48 hours.

  • Hosting in our data centre

    No server room? We place the server in our data centre, take care of power, cooling and connectivity, and you access it remotely.

How we work on AI servers

Description

AI server PC Praha Turin Dual is based on the dual-socket ASRock TURIN2D24XGM board with two AMD EPYC 9355 processors (2× 32 cores, Zen 5). Twenty-four DDR5 ECC slots allow 768 GB or 1.5 TB of system memory. It is designed for research, model fine-tuning and running the largest open models for the whole company – and, in the version with sixteen RTX Pro 6000 cards, for workloads that would otherwise require renting a cluster.

Configuration

Processors 2× AMD EPYC 9355 (2× 32 cores, Zen 5, SP5 socket)
Motherboard ASRock TURIN2D24XGM (dual-socket)
Graphics cards 4, 6 or 8× of your choice: RTX 5080, RTX 4090, RTX 5090, Intel Arc Pro B70, NVIDIA L4, NVIDIA L40 or RTX Pro 6000; or 16× RTX Pro 6000 (PCIe 5.0 x8)
Memory 768 GB or 1.5 TB DDR5 ECC
Storage NVMe 4 or 8 TB
Power supply redundant server power supplies (double set for 16 cards)
Chassis 7U rack

How much graphics memory you need

Graphics memory Example configuration What runs comfortably
64–192 GB 4–8× RTX 5080 / RTX 4090, 4–6× RTX 5090, 4× L40 Llama 3.3 70B up to GLM-4.6 355B (Q4)
256–384 GB 8× RTX 5090, 6–8× L40, 4× RTX Pro 6000 DeepSeek-V3.2 / R1 671B (Q2), Qwen3-Coder-480B, Kimi-K2 1T (1.58-bit); full fine-tuning of a 70B model
576–768 GB 6–8× RTX Pro 6000 the largest open models with headroom for context and many users
1,536 GB 16× RTX Pro 6000 further training of large models (100B+) and running several of the largest models at once

Model sizes are indicative (Q4 quantisation unless stated otherwise), based on our configurator. Actual memory requirements depend on context length and the number of concurrent users. For language models the memory of the cards adds up; image and video generation needs the whole model on a single card.

What you get

  • A server built from new components – we assemble and test it in Prague.
  • Pre-installed software for whatever you want to run on the server (e.g. Ubuntu, drivers and CUDA, Docker, vLLM, llama.cpp, Ollama, ComfyUI). It is ready to work as soon as you switch it on.
  • 72 hours of stress testing under full load before handover.
  • 24-month warranty. We keep spare parts in Prague and provide service intervention within 24/48 hours.
  • The server does not have to stand at your premises – you can keep it in our data centre (hosting) and we take care of power, cooling and connectivity.

Operation and installation

The server belongs in a server room. With eight cards it draws roughly 1.2 to 5.4 kW, with sixteen RTX Pro 6000 cards around 10 kW – we will design the power and cooling with you, or run the server in our data centre. We can add fast networking of up to 400 GbE on request.

Frequently asked questions

When is Turin Dual worth it instead of Genoa?
When you need more than 768 GB of system memory, more cores for data preparation or 16 cards in one server. For 4–8 cards with 768 GB of RAM, Genoa is cheaper.
How do 16 cards fit into one server?
The 16-card version uses only RTX Pro 6000 cards connected via PCIe 5.0 x8 and a double set of server power supplies. We will confirm the exact configuration and delivery date in the quote.
Is it suitable for training?
Yes – it handles full fine-tuning of 70B models with eight or more RTX 5090 cards, and with 16× RTX Pro 6000 even further training of models with over 100 billion parameters.
What do I need to run it?
A server room with power supply and cooling for several kW up to roughly 10 kW. If you do not have one, we run the server in our data centre and you access it remotely.

Customer reviews

There are no reviews yet.

Only customers who bought the product from us can leave a review – we send them a review link by e-mail after the purchase.

More AI servers

View all