AMD第五代EPYC为谷歌云TPU 8tx提供主机算力：首款带硬件级机密计算的云端TPU系统上线
AMD 11月9日宣布，第五代 AMD EPYC 处理器正为谷歌云（Google Cloud）新推出的 TPU 8tx 提供主机基础设施，帮助企业在云端运行最敏感的超级智能（SI）工作负载时保护数据与模型安全。TPU 8tx 是首款交付硬件级机密计算的云端张量处理单元（TPU）系统。
AMD 说，这次扩大合作把 EPYC 处理器的性能、AMD Infinity Guard 安全功能与 x86 软件兼容性，同谷歌云 TPU 的加速能力结合起来，企业可以据此部署推理、微调与强化学习等工作负载。谷歌云在 TPU 8tx 上采用的机密计算架构由第五代 EPYC 主机处理器支持，机构可以在云端大规模运行敏感工作负载，不必在数据保护上让步。
谷歌同日在其官方博客介绍，TPU 8tx 是谷歌第八代 TPU 家族的第三款系统，专为安全、机密的推理优化，也适合小规模训练，面向运行前沿或自有模型与智能体的受监管企业和政府机构。它与 TPU 8t 采用同一款芯片，搭配第五代 AMD EPYC 主机，把安全直接建进硬件层，让敏感数据与模型权重在正在使用时也保持受保护和隔离状态。
谷歌还说，TPU 8tx 采用为实时推理打造的低延迟网络，与 TPU 8t 相比延迟最高可降低约一半，性能代价接近于零；团队可以在同一套系统上用私有数据微调模型、把大模型蒸馏成小模型，并通过强化学习训练智能体，不必把敏感数据在不同集群之间迁移。
AMD 总部位于加利福尼亚州 Santa Clara，其 EPYC 是面向服务器与数据中心的处理器产品线。谷歌同日还发布了第八代 TPU 家族的第四款系统 TPU 8i Max，主打超大上下文窗口与单机架最高计算密度。
（来源：AMD）
5th Gen AMD EPYC Powers Google Cloud TPU 8tx, the First Cloud TPU With Hardware-Enforced Confidential Computing
Advanced Micro Devices (AMD) announced on November 9 that 5th Gen AMD EPYC processors are powering the host infrastructure for Google Cloud TPU 8tx, the first cloud Tensor Processing Unit (TPU) system to deliver hardware-enforced confidential computing, helping enterprises protect sensitive super intelligence (SI) models and data in the cloud.
AMD said the expanded collaboration combines EPYC CPU performance, AMD Infinity Guard security capabilities and x86 software compatibility with Google Cloud TPU acceleration, so enterprises can deploy inference, fine-tuning and reinforcement learning workloads. Google Cloud's confidential computing architecture on TPU 8tx is supported by 5th Gen AMD EPYC host processors, letting organizations run their most sensitive workloads at scale in the cloud without compromising on data protection.
In its own introduction the same day, Google said TPU 8tx, the third system in its eighth-generation TPU family, is optimized for secure and confidential inference and is also suited to small-scale training, built for regulated enterprises and governments running frontier or proprietary models and autonomous agents alongside their most sensitive data. Built on the same silicon as TPU 8t and paired with 5th Gen AMD EPYC CPU hosts, TPU 8tx builds security directly into the hardware layer, keeping sensitive data and model weights protected and isolated even while in active use.
Google added that TPU 8tx uses the same low-latency network created for real-time reasoning, delivering up to 2x lower latency compared to TPU 8t with a near-zero performance penalty, and that teams can fine-tune models using private data, distill large models into smaller ones and train agents through reinforcement learning on a single system, without migrating sensitive data between separate clusters.
AMD is headquartered in Santa Clara, California, and EPYC is its processor line for servers and data centers. Google also introduced TPU 8i Max the same day, a fourth eighth-generation system built for massive context windows and the highest compute density per rack.
(Source: Advanced Micro Devices, Inc)
来源：AMD
发布：GSM·美国（发布链接）
图片来源：图：谷歌位于爱荷华州 Council Bluffs 的数据中心。Chad Davis 摄，CC BY 2.0，经 Wikimedia Commons。 / Photo: Google's data center in Council Bluffs, Iowa. By Chad Davis, CC BY 2.0, via Wikimedia Commons.
GSM Global · 本篇英文版 English edition of this story
