The hardware powering an AI agent determines how smart it is. To drive AI, companies must buy the most powerful chips available. As a result of the AI boom, NVIDIA has seen a huge surge in popularity. However, other firms are also developing their own artificial intelligence chips.
It has been years since Google began developing its artificial intelligence chips. Google’s Pixel devices will sound pretty familiar if you are familiar with the company’s processing units, which are called TPUS (Tensor Processing Units). Google’s AI relies on them. Google releases new AI chips every few years, but its latest model represents an exponential improvement.
Despite numerous hiccups, Google’s AI is among the best in the world. This is due to the models’ incredible power. Ironwood is the company’s 7th-generation AI chip architecture. It will power Google’s AI tech going forward. With these specs, we can expect Google’s AI to improve dramatically.
Google released its TPU v5p chip in 2023. This chip has 9860 chips in a pod, while Iron has 9216 chips in a pod. As a result, overall capabilities increase dramatically. The Ironwood has 192GB of memory, which is a 102% increase over TPU v5p (95GB). The combined TFLOPS of Ironwood and TPU v5p are more than 10 times as high, with 4.614 Exaflops compared to 459 TFLOPS for TPU v5p.
Google did not compare this new chip with the 2024 Trilium chip. Ironwood is more like TPU v5p than Trilium, and the company said Trilium was a follow-up to TPU 5pe.
