Compute
For Platform 2.0 Only
GPU Server
병렬 처리에 최적화된 GPU 서버의 고성능 컴퓨팅 파워를 제공합니다.
The Best Choice for High Performance Computing, Such as Deep Learning, AI, or Autonomous Driving
A Graphics Processing Unit (GPU) is a computing resource optimized for high performance, simultaneous processing. Flexible response in dynamic business environments is possible by accessing the desired data promptly and accurately through GPU parallel computing. It is most appropriate for large operations, such as scientific computing or data analysis.
- High Performance GPU Environment that can be Utilized in Various Fields
- GPU is applicable to various fields, such as image processing, rendering, scientific computing, and machine learning. A fast and stable computing environment is provided with the exclusive optimized configuration of NAVER CLOUD PLATFORM on the GPU in which the latest TESLA architecture is applied.
- Simple Installation and Configuration
- The most recent GPU driver and NVIDIA CUDA package (development toolkit), which are essential for the configuration, are provided by default. This allows convenient configuration and use of the GPU environment.
Built-In Package: NVIDIA CUDA Toolkit (CUDA v9.0) & NVIDIA CUDA Library (cuDNN v7.4.1.5)
- Easy Scalability
- GPU resources can be increased or decreased freely when creating a server or as needed.
(Maximum of 2 GPUs per VM) Also, the default storage space can be increased by simply adding additional capacity.
Detailed Features
Various features are provided to improve the stability and convenience of the GPU server.
Providing an Outstanding GPU Performance with Scalability
Through the optimized configuration environment provided by NAVER CLOUD PLATFORM, settings are uncomplicated.
You can easily add GPU resources via the console, if necessary. (Maximum of 2 GPUs per VM)
Multi-GPU Feature (T.B.D.)
A maximum of 8 GPU resources are provided per Virtual Machine (VM). Performance can be freely adjusted at the times you designate.
Faster results in a shorter period of time can be obtained with this higher computing power.
Specifications of Provided GPU
| P40 | |
|---|---|
| # of CUDA Cores | 3,840 |
| Peak Single Precision | 12 TeraFLOPS |
| Peak INT8 | 47 TOPS |
| Low Precision | 4x 8-bit vector dot product with 32-bit accumulate |
| Video Engines | 1x decode engine, 2x encode engines |
| GDDR5 Memory | 24 GB @ 346 GB/s |
| Power | 250W |
Related Services
Pricing information
For flexibility, we provide both time and usage-based pricing models
We provide a GPU server with the latest NVIDIA TESLA P40 model.
- The basic disk for installing the OS is provided. You can create as many additional disks as supported by the OS (up to 2,000GB per disk).
- Servers can be used on an hourly or monthly basis. If you use less than one month of storage, the usage fee is calculated based on days used.
- Only available in the Korea region.
| Type | Available Specifications | Usage Fee | |||||
|---|---|---|---|---|---|---|---|
| GPU | GPU Memory | vCPU | Memory | Disk | Hourly | Monthly | |
| GPU | 1 EA | 24GB | 4 EA | 30GB | 50GB (HDD) | 1,667 KRW | 1,200,000 KRW |
| 50GB (SSD) | 1,671 KRW | 1,202,880 KRW | |||||
| 2 EA | 48GB | 8 EA | 60GB | 50GB (HDD) | 3,334 KRW | 2,400,000 KRW | |
| 50GB (SSD) | 3,338 KRW | 2,402,880 KRW | |||||
Network Fee
Outbound traffic usage is billed based on the following pricing table.
| Classification | Networks | Unit | Usage Level | Meter Rate (monthly) |
|---|---|---|---|---|
| Inbound Traffic | All | GB | Same for all sections | Free |
| Outbound Traffic | Traffic through the Internet (including traffic toward another region via public IP) | GB | 20GB or Less | Free |
| 20GB–5TB | 100 KRW | |||
| 5TB–10TB | 90 KRW | |||
| 10TB–30TB | 80 KRW | |||
| Over 30TB | 70 KRW | |||
| Traffic between different zones via a public IP | GB | Same for all sections | 10 KRW | |
| Traffic within the same zone via a public IP | GB | Same for all sections | 10 KRW | |
| Traffic between different zones via a private IP | GB | Same for all sections | 10 KRW | |
| Traffic within the same zone via a private IP | GB | Same for all sections | Free |
- As of Sep. 1, 2018, new free usage tier will be applied to new subscribers (1TB -> 20GB).
Provided OS
- The basic disk for installing the OS is provided. You can create as many additional disks as supported by the OS (up to 2,000GB per disk).
| OS Classification | Usage Fee | |||
|---|---|---|---|---|
| OS | Version | 32/64-bit | Hourly | Monthly |
| Linux | CentOS 7.3 | 64bit | Free | Free |
| Ubuntu 16.04 | 64bit | Free | Free | |
Pricing Information for Server Service Stoppage
- When stopping a server service, only the fee on the basic disk provided for the installation of OS on some server types will be charged.
- Fees for other resources connected to the server, such as public IP, load balancer, and additional disks will be
charged regardless of server stoppage.
Please be sure to return the resources that are not used to avoid unnecessary billing problems.
(Fees on OS and MSSQL will not be charged.) - Servers with discounted rates at the time of server stoppage may not stop the server for more than 90 days at a time or for more than a cumulative 180 days within 12 months. Servers that have exceeded the possible period for stopping the server will be withdrawn after notifying the customer. At this time, the customer data stored in the server will be deleted after being backed up and stored by the company for 30 days and the customer can request to reuse the server or recover data only within the storage period. (Details above shall be effective starting August 1st, 2017 when the terms and conditions will be revised, and the suspension period will be calculated based on August 1st, 2017.)
| Classification | Target Server Type |
|---|---|
| Applicable for Discounts at Server Stoppage | Compact Server and Standard Server |
| Not Applicable for Discounts at Server Stoppage | High Memory Server, Micro Server, CPU Intensive Server, Virtual Dedicated Server, GPU Server and Baremetal Server |
Was this page helpful?