Accelerated Server (GPU/NPU)

Get Started

Virtual AI Accelerator Computing Service for AI Training and Inference

Accelerated Server is a service that provides high-performance virtual servers based on AI accelerator resources such as GPUs and NPUs. It provides a high-performance, high-efficiency infrastructure environment optimized from training to large-scale inference for AI model development, and is available as much as you need when you need it.

Comparison of Accelerated Server Types

GPUs handle both graphics and large-scale general computing simultaneously, whereas NPUs are dedicated processors engineered to run AI computations with speed and efficiency.
The key differences and features of the two processors are follows.

Classification of Accelerator service features, and a detailed explanation table describing GPUs and NPUs
Category	GPU	NPU
Definition	Graphics Processing Unit	Neural Processing Unit
Key Role	Graphics rendering, large-scale data parallel processing	AI training and inference (Deep learning algorithms)
Processing Method	Processes identical operations simultaneously using thousands of cores	Optimized for sequential AI algorithm execution (deep learning) mimicking the human brain
Features	Ideal for gaming, complex graphics rendering, and massive AI model training	Fast and efficient AI prediction (Inference), Low power consumption
Examples	Running High-end games, Large-scale AI model training	Smartphone facial recognition, Real-time translation, and Autonomous driving

GPU

Definition Graphics Processing Unit
Key Role Graphics rendering, large-scale data parallel processing
Processing Method Processes identical operations simultaneously using thousands of cores
Features Ideal for gaming, complex graphics rendering, and massive AI model training
Examples Running High-end games, Large-scale AI model training

NPU

Definition Neural Processing Unit
Key Role AI training and inference (Deep learning algorithms)
Processing Method Optimized for sequential AI algorithm execution (deep learning) mimicking the human brain
Features Fast and efficient AI prediction (Inference), Low power consumption
Examples Smartphone facial recognition, Real-time translation, and Autonomous driving

Overview

Easy Computing Environment
It is easy to provision, monitor, and check usage of Accelerated Server using a web-based console.
Various Specs of Services
Depending on the type and size of AI workloads, users can select different accelerator types and onfigurations, such as GPUs and NPUs. Accelerated Server provides a high-performance computing environment optimized for AI workloads.
Strong Security
Accelerated Servers are securely protected by inbound/outbound traffic control with external Internet or other VPC through Security Group service. Real-time monitoring also support reliable operation of computing resources.
Easy Network Settings
The initial setting for subnet/IP of Accelerated Server can be changed easily. NAT IP can be used/turned off based on the needs of the user, providing easy network connection.

01

04

Service Architecture

Key Features

Server creation/management
1. Select GPU or NPU card based on task type and required specifications(B300, H100, A100 and RNGD)
2. Offer various OS standard images for creating resources conveniently(RHEL and Ubuntu)
3. Securely connect to the OS using key pairs
4. Manage tag creation/revision
Storage and network connection
1. Provide additional storage besides OS disk
2. Set up subnet/IP and enable the use/turn off NAT IP setting
3. Provide Security Group integration setting
  ※ Custom images do not support backup services
Convenient service management
1. Manage creation or deletion of custom images
2. Manage custom image copies between accounts

Let’s talk

Whether you’re looking for a specific business solution or just need some questions answered, we’re here to help

Get in Touch