Google unveiled its next generation of Tensor Processing Units (TPU), Trillium, at its annual Google I/O developer conference. The TPUs are scheduled for release later this year.
Google's Trillium boasts several enhancements from the previous generation. This includes a 4.7x increase in compute performance per chip, wider matrix multiply units (MXUs), higher clock speed, and double the memory bandwidth. The new chips are also 67% more energy-efficient than the previous generation.
The chips uniquely carry the third generation of SparseCore, an accelerator specially designed for handling large-scale embeddings often found in advanced recommendation or ranking tasks. This enables Trillium TPUs to train models faster and serve them with lower latency.
By using this site, you agree to allow SPEEDA Edge and our partners to use cookies for analytics and personalization. Visit our privacy policy for more information about our data collection practices.