---
title: Gree Implements Google Cloud's AMD EPYC C4D Instances, Achieves Over 25% Cost Efficiency
url: https://www.dataloco.com/en/gree-implements-google-clouds-amd-epyc-c4d-instances-achieves-over-25-cost-efficiency
published: 2026-09-22T21:20:32+00:00
language: en
section: Cloud
source: https://ascii.jp/elem/000/004/436/4436512/?rss
organizations: amd, c4d, epyc, google
publisher: Dataloco
---

# Gree Implements Google Cloud's AMD EPYC C4D Instances, Achieves Over 25% Cost Efficiency

Gree has announced the implementation of AMD EPYC processors in their smartphone game applications. The company has adopted Google Cloud's fifth-generation AMD EPYC processors via C4D instances, reportedly achieving over 25% higher cost performance compared to their previous environment.

Gree operates multiple large game titles with hundreds of thousands of daily active users (DAU), necessitating infrastructure with high CPU performance and flexible scalability to accommodate fluctuations in user access during game events and peak hours. Their game servers utilize a container-based architecture, managing hundreds of game instances that utilize more than 16 cores per title.

Typically, the introduction of new CPUs or instance types requires extensive validation processes. However, Gree opted not to launch a dedicated validation project but instead integrated evaluation procedures into their existing game development processes.

During the development phase, Gree established a setup where game servers operate across multiple instance types, allowing developers to progress in feature development without considering the differences in CPUs.

In the load testing phase, they replicated communication patterns based on actual user access to compare metrics such as latency and CPU usage. The testing confirmed the superior performance and cost efficiency of the C4D instances.

In the QA and production deployment phase, Gree conducted quality verification (QA) using the selected instance types. In the production environment, they achieved a balance between resource allocation and high availability by combining multiple instance types with Spot instances.

This implementation has enabled Gree to run a high-density infrastructure environment with fewer servers, executing numerous container workloads efficiently. Load testing demonstrated improvements in latency, reductions in CPU usage, and overall cost performance enhancement of over 25% (as evaluated by Gree).

## This story in other languages

- [Türkçe](https://www.dataloco.com/tr/gree-google-cloud-uzerinde-amd-epyc-islemcili-c4d-orneklerini-kullanmaya-basladi)