GREE Adopts AMD EPYC C4D Instances on Google Cloud for 25% Cost Performance Gain
GREE's numbers give AMD and Google Cloud a named game-server reference for the C4D instance type, and the rollout shows the verification process, not the silicon, is the part that normally blocks a CPU migration.
Reporting from 1 source: ASCII.jp.
AMD published a case study of GREE's use of AMD EPYC processors in its smartphone game operations. GREE adopted C4D instances with 5th Gen AMD EPYC on Google Cloud and measured more than 25% higher cost performance than its previous environment for the target workloads, a figure GREE evaluated. The company skipped a dedicated verification project and folded the evaluation into its existing game development workflow, covering development, load testing, and QA.
GREE's game servers run on a container-based architecture, with instances using 16 or more cores per title across hundreds of instances, and the titles carry daily active users in the hundreds of thousands. Introducing a new CPU or instance type normally means a large amount of verification work. GREE did not start a dedicated project. It built an environment where game servers run on multiple instance types, so developers could work without tracking CPU differences, then compared latency and CPU usage under load that reproduced actual user access. Production mixes multiple instance types and Spot instances. The company said the result is a denser infrastructure running many container workloads on fewer servers.
Synthesized by Yomimono from the 1 cited source below, including Japanese-language reporting where cited, then editorially reviewed before publishing.