Cloudflare’s Gen 13 servers double our compute throughput by rethinking the balance between cache and cores. Moving to high-core-count AMD EPYC ™ Turin CPUs, we traded large L3 cache for raw compute density. By running our new Rust-based FL2 stack, we completely mitigated the latency penalty to unlock twice the performance.

Cloudflare's platform is a leading provider of internet security and performance solutions, offering insights into web security, content delivery, and DNS management. Through documentation, blog posts, and webinars, Cloudflare provides insights into protecting websites and applications from cyber threats and improving performance. Developers and IT professionals can learn about CDN (Content Delivery Network), DDoS mitigation, and firewall configurations to secure and accelerate web traffic.

Cloudflare

Cloudflare's Gen 13 server launch details how they doubled edge compute throughput by switching from AMD EPYC Genoa-X (with large 3D V-Cache) to high-core-count AMD EPYC Turin 9965 processors. The key challenge was that Turin's 192 cores share far less L3 cache per core (2MB vs 12MB), causing severe latency regressions under their legacy NGINX/LuaJIT-based FL1 stack. Hardware tuning alone couldn't solve the problem. The solution was FL2, a complete Rust rewrite of Cloudflare's core request handling layer built on Pingora and Oxy frameworks. FL2's leaner memory access patterns eliminated cache dependency, cutting latency penalties by 70% and enabling linear throughput scaling with core count. The result: 2x throughput vs Gen 12, 50% better performance per watt, and 60% higher rack throughput — all while meeting latency SLAs.

Launching Cloudflare’s Gen 13 servers: trading cache for cores for 2x edge compute performance

Diagnosing the problem with performance counters

Incremental gains with performance tuning

The opportunity: FL2 was already in progress