Urgent.News

What's breaking now, across hundreds of outlets.

Tech

TPUs

Google has expanded its eighth-generation Tensor Processing Unit (TPU) portfolio with two new systems: TPU 8tx for secure and confidential inference and TPU 8i Max for massive context windows and high compute density per rack. These additions aim to address the evolving needs of businesses working with increasingly complex AI models and regulated workloads.

TPU 8tx is optimized for secure and confidential inference, making it ideal for regulated enterprises and governments using sensitive data in their autonomous agents. With hardware-level data security built directly into the system, TPU 8tx keeps sensitive data protected and isolated while in active use, ensuring that even while being processed, data remains secure.

This system, built on the same silicon as TPU 8t and paired with 5th Generation AMD EPYC CPU hosts, also offers a fast inference network, resulting in up to 2x lower latency compared to TPU 8t with a negligible performance penalty.

On the other hand, TPU 8i Max is designed to tackle the memory wall issue that arises when multiple AI agents collaborate on tasks such as large-scale training, real-time reasoning, or processing vast amounts of data. To address this, Google has combined two low-latency TPU 8i chips onto a single package, doubling the compute, SRAM, and HBM bandwidth compared to TPU 8i. This new system allows processors to remain equipped with essential data while handling the growing memory demands of complex AI workloads.

Rewritten by urgent.news; may contain errors — check the original.

Read the original at blog.google →

More in Tech

More from Monday 9 November →