top of page
Technology

Faster,
Less power,
Higher Cost-efficiency.

Proven Tech
in diverse commercial applications providing faster AI/DL processing with higher power-cost efficiency.

And it comes in forms of chip, card, and server as an integrated
hardware-software solution for "Cloud AI as a Service" at datacenters.
chip.png

AI Processor

card.png

PCIe Card

server.png

Server

cloud.png

Datacenter (Cloud)

State-of-the-art Tech

Technology and architecture advantages to meet all the requirements the Inference market expects

Satisfying performance in low latency with

wide flexibility

Delivering best inference performance

Power-Cost Efficiency

Delivering the best AI inference performance with high efficiency in between competitors.

Save up to 50% cost out of the competitive

GPU tiers

4x effectiveness for Performance to cost and Watt

Field-Proven Accomplishment

Commercial AI-based field applications are currently running on SAPEON.

Achievement for both performance and cost

innovation

Running on diverse commercial applications

Excellent R&D Manpower

Over 95% of our R&D members are MS or higher of related field. With our great manpower, we're aiming only SUPEX goals.

X220

Korea's first AI processor for datacenters.

SAPEON has an optimal architecture for low-latency, large-scale inference of deep neural networks. Our products are optimally designed to process artificial intelligence tasks faster, using less power by efficiently processing large amounts of data simultaneously.
Get continuous, unified and high-quality AI services across the devices with SAPEON.

2020

Commercial Sample Released

X220Ch_1ip.png

AI Inference

2023

AI Inference

AI Training

X330.png
X340.png
X350.png

2025

Chip_back
X430.png

AI Inference

AI Training

HBM3

X220 Specification

X220-Compact
X220-Enterprise
SAPEON-X220-Enterprise-Card.webp
SAPEON-X220-Compact-Card.webp

Precision

Integer 16, 8, 4 bit

Integer 16, 8, 4 bit

Int8 Performance

87 Tera OPS

(Boost Mode 106 Tera OPS)

174 Tera OPS

(Boost Mode 213 Tera OPS)

Effective Performance

(Resnet-50, Batch 18)

6,785 FPS

13,570 FPS

Memory Capacity

8 GB

16 GB

Host Interface

PCIe Gen3 x 16 Lane

PCIe Gen3 x 16 Lane

Max Power Consumption

65 W

135 W

SAPEON X220  is SAPEON's self-developed AI chip that aims for high-efficiency large-scale cloud inference. Its deep learning computation speed is 6.7 kilo-frames per second, meaning it can process 6,700  images per second.

SAPEON X330  is scheduled to be released in 2023. If the existing SAPEON X220 was designed to optimize a semiconductor’s inference skill, the SAPEON X330 will include not only inference but also new learning features, such as the ability to learn data, that enhance the AI processing in real-time. 

Demo

Application Demo

Image Classification

In the image classification demo, the performance of SAPEON and GPU is compared. During ResNet-50, the most commonly used artificial neural network  is running, SAPEON X220 shows superior performance compared with the GPU. 

Object Detection

In the Object Detection demonstration, the task is to recognize people, cars, bicycles, and so on from video footage as it is played. The faster these objects are detected, the faster the video gets played. YOLO-v3, the most famous artificial neural network for object detection, is being used for comparison. SAPEON shows better performance but lower power consumption. 

Image Quality Enhancement

SUPERNOVA

This demonstration shows the performance of SAPEON applied to SUPERNOVA, which is SKT's technology to upscale low-resolution images using AI. To change image resolution in real-time, the AI processing speed must be fast. In the demonstration, SAPEON shows faster speed for improving resolution compared with a GPU. Not only that but it is possible to produce the result you want at a lower cost and consume less power.

5G Edge Cloud

SAPEON operating on MEC (Network Edge) via the 5G network. The application program for detecting objects is running. The artificial neural network that cannot operate on a mobile phone is being remotely executed on SAPEON through the 5G MEC. 

VPS
(Visual Positioning System)

SAPEON running VPS, an indoor navigation technology. Using SAPEON makes it possible to quickly and accurately locate one's position, and it enables navigation technology to work properly.

SAPEON's application demo uses BERT, a natural language processing model, to perform tasks that answer questions based on input. X220 is showing superiority compared to GPU, not only in Vision but also in the Language field.

SAPEON
Use Cases

frame.png

Commercial AI-based field applications are currently running on SAPEON. SAPEON X220 has been applied to AI assistant service, "NUGU," as well as a variety of services of our affiliates such as media content selection and media resolution upgrading service, "Supernova," and cloud-based video surveillance service, "T View". We provide integrated AI solutions such as AI infrastructure, software, and applications in various areas. 
 

1

Mobility & Robot

As of late, the use of AI robots has emerged in various applications, including directional guidance delivery, and disease control and prevention. VLAM(Visual Localization And Mapping) and vision AI technologies are utilized in order to provide accurate location information to the cloud robot and to analyze collected videos. SAPEON allows for more efficient and economical AI-based robot services.

2

Media

SAPEON is utilized in SK telecom media’s quality advancement technology. SUPERNOVA, which enhances low-quality, low-resolution videos with AI technology. Supernova, running on SAPEON, can not only improve the video quality significantly, but also enables large scale services, providing universal AI services to wider public. 

3

Safety & Security

SK shieldus applied SAPEON of its AI-based video surveillance system that detects abnormal movements such as intrusion and falling. It further contributes to ensuring safety through applications including the detection of driver negligence, surveillance of unlawful train boarding, and general safety inspection. 

4

Smart Factory

SK shieldus applied SAPEON of its AI-based video surveillance system that detects abnormal movements such as intrusion and falling. It further contributes to ensuring safety through applications including the detection of driver negligence, surveillance of unlawful train boarding, and general safety inspection. 

5

Medical

Utilization of Vision AI technology in the medical field is rapidly increasing for purposes including early cancer detection and epidemiological analysis. The utilization of SAPEON in high-quality, affordable medical services contributes to a healthier, more prosperous society.

Use cases
bottom of page