Cloudflare has made waves in the tech world with the launch of its 12th-generation servers, now powered by AMD’s EPYC 9684X Genoa-X processors.
These powerful servers bring significant upgrades in terms of both performance and efficiency, with Cloudflare promising up to 145% more requests per second (RPS) and 63% better power efficiency compared to their previous Gen 11 models.
These advancements are crucial for handling growing data traffic, AI workloads, and computationally demanding applications.
The Power of AMD EPYC Genoa-X: 96 Cores and Massive Cache
At the heart of Cloudflare’s new servers is the AMD EPYC 9684X Genoa-X CPU, which comes with a whopping 96 cores, 192 threads, and an industry-leading 1152MB of L3 cache.
This huge cache is three times larger than the standard AMD Genoa processors, enabling the server to handle data-intensive tasks with much lower latency.
Cloudflare reports that Genoa-X CPUs deliver a 22.5% improvement in performance compared to other models in the EPYC lineup, making them ideal for high-demand environments.
Enhanced Efficiency and Optimized Design
Despite the leap in power, Cloudflare hasn’t sacrificed efficiency. The new servers are designed to be more power-conscious, thanks to several innovations.
The Gen 12 servers utilize a 2U form factor, an upgrade from the previous 1U size, allowing for better airflow and cooling. This change alone has reduced fan power consumption by 150W, significantly lowering overall energy use.
While the Gen 12 servers consume about 600W during typical operations—up from 400W in Gen 11—the performance gains justify this increase.
Cloudflare’s enhanced thermal-mechanical design and the shift to dual 800W Titanium-grade power supplies help maintain energy efficiency, contributing to a 63% better overall power efficiency compared to previous generations.
AI Workload-Ready with GPU Scalability
Cloudflare’s new servers are also designed to support growing AI demands. With 384GB of DDR5-4800 memory across 12 channels, 16TB of NVMe storage, and dual 25 GbE network connectivity, the Gen 12 servers are equipped for high-memory throughput and faster data processing.
The architecture allows for GPU scalability, supporting up to two PCIe add-in cards, enabling Cloudflare to manage AI inference tasks more efficiently.
This makes the servers ideal for regions that have high demand for AI processing, reducing latency and boosting performance for tasks like machine learning and AI model deployment.
Looking forward, Cloudflare has already begun testing 5th generation AMD EPYC “Turin” CPUs for its next generation, Gen 13 servers, ensuring they stay ahead in performance and efficiency.
Security Upgrades: Root of Trust and DC-SCM 2.0
On the security front, Cloudflare has integrated advanced features into the Gen 12 servers, including hardware root of trust (HRoT) and Data Center Secure Control Module (DC-SCM 2.0).
These technologies ensure the integrity of boot firmware and provide modular security, protecting the servers from firmware-level attacks and reducing vulnerabilities in their global data centers.
Cloudflare’s AI Developer Product Enhancements
Alongside its hardware updates, Cloudflare has made significant improvements to its AI developer tools.
Workers AI, now powered by stronger GPUs, can handle larger models such as Meta’s Llama 3.1 70B and Llama 3.2, enabling more complex AI tasks to be managed efficiently across Cloudflare’s 180+ city network.
Cloudflare has also upgraded its AI Gateway, a tool that helps developers monitor and optimize AI deployments.
With new persistent logs (currently in beta), AI Gateway now supports in-depth performance analysis through features like search, tagging, and annotation, allowing developers to fine-tune their AI models with greater precision.
Additionally, Cloudflare’s vector database, Vectorize, has reached general availability. It now supports indexes of up to five million vectors, helping to reduce latency and speed up AI processes.
A simplified unit-based pricing structure has been introduced across Cloudflare’s AI products—Workers AI, AI Gateway, and Vectorize—offering better transparency and cost management for developers.
Cloudflare’s 12th-generation servers represent a significant leap forward, not just in terms of raw power but also in efficiency and security.
