發生什麼事
Google Cloud's announcement brings the eighth-generation range to four systems, with TPU 8i Max and TPU 8tx in the third and fourth spots. TPU 8tx targets confidential computing; TPU 8i Max offers extra-large context windows and leads the lineup in per-rack compute density. TPU 8i serves low-latency inference, while TPU 8t supports large-scale training. 🗣
01Sources
- @googlecloud (publisher: Google Cloud official account)
- Evidence: 🗣 Official account statement