Skip to product information
GIGABYTE G894-SD3 – official product photo (1/4)
  • GIGABYTE G894-SD3 – official product photo (1/4)
  • GIGABYTE G894-SD3 – official product photo (2/4)
  • GIGABYTE G894-SD3 – official product photo (3/4)
  • GIGABYTE G894-SD3 – official product photo (4/4)
GIGABYTE G894-SD3 – 8× NVIDIA B300 (HGX), 2,160 GB HBM3e, 2,048 GB RAM, 8U

GIGABYTE · AI and GPU servers

GIGABYTE G894-SD3 – 8× NVIDIA B300 (HGX), 2,160 GB HBM3e, 2,048 GB RAM, 8U

Key specifications

Graphics cards
8× B300 (HGX)
Graphics memory
2160 GB
Processor
2× Intel Xeon 6776P
Memory
2048 GB DDR5
Storage
2× 7,68 TB NVMe SSD
Indicative price 13 920 570 Kč excl. VAT
16 843 890 Kč incl. 21% VAT. We confirm the final price, configuration and delivery date in our quote.
  • Lead time: Na zakázku – termín po konzultaci
  • We usually send a no-obligation quote within 2 working days.
  • No server room? We can host it in our data centre.
Request a quote

No-obligation enquiry

Built individually to order – we will prepare the configuration, price and delivery date after a consultation.
8× NVIDIA B300 Blackwell Ultra SXM, 2,160 GB HBM3e, 2× Xeon 6776P, 2,048 GB DDR5, 2× 7.68 TB NVMe, 8U rack.
  • 24-month warranty, claims handled directly by us
  • We help with configuration and software
  • We can run it in our data centre on request

Built individually to order – we will prepare the configuration, price and delivery date after a consultation.
8× NVIDIA B300 Blackwell Ultra SXM, 2,160 GB HBM3e, 2× Xeon 6776P, 2,048 GB DDR5, 2× 7.68 TB NVMe, 8U rack.

Full description, video and specifications

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.

  • Warranty and service through us

    We handle warranty claims and communication with the manufacturer for you, and help with service and setup right here in Prague.

  • 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

GIGABYTE G894-SD3 with the NVIDIA HGX B300 platform in its full configuration: 2 TB of system memory and 2× 7.68 TB NVMe for both training and inference.

What it is good for

A server for training and fine-tuning large models and for production inference serving hundreds of users. 2,160 GB of HBM3e memory holds even the largest open models at full or 8-bit precision – for example DeepSeek-V3/R1 (671 billion parameters) or Kimi K2 (one trillion parameters) – with headroom for long context and many concurrent requests. The accelerators are interconnected via NVLink.

The configuration with 2 TB of RAM and 2× 7.68 TB NVMe gives you headroom for training and fine-tuning, data pre-processing and a local model cache directly on the server.

This server belongs in a data centre: it is 8U high and draws more than 10 kW at full load. If you do not have your own server room, we will place it in our hosting in the data centre.

Manufacturer specifications

Platform NVIDIA HGX B300, 8× Blackwell Ultra SXM
GPU memory 2,160 GB HBM3e (8× 270 GB)
Processors 2× Intel Xeon 6776P
System memory 2,048 GB DDR5
Storage 2× 7.68 TB NVMe SSD
Form factor 8U rack
Operating system Ubuntu Linux

Buying from us

  • Sold to businesses (B2B), made to order. The goods are ordered specifically for you – a confirmed order cannot be cancelled and the goods cannot be returned.
  • We confirm the delivery date and final configuration in our quotation.
  • Warranty according to the manufacturer's terms; we handle servicing and communication with the manufacturer.
  • We can prepare the software (Ubuntu, NVIDIA drivers, CUDA, vLLM, Docker) and place the server in our hosting in the data centre.

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