Google Looks to Compete with NVIDIA
Google looks to dominate at the silicon level.
Created on April 6|Last edited on April 6
Comment
Google has published details about its artificial intelligence (AI) supercomputer, asserting that it is faster and more efficient than competing systems from NVIDIA. As machine learning models continue to advance, companies are racing to develop more powerful and efficient AI chips and systems.
Google vs. NVIDIA
NVIDIA currently dominates the market for AI model training and deployment with a 90% market share [1]. However, Google has been designing and deploying AI chips called Tensor Processing Units (TPUs) since 2016. Despite this, some believe Google has fallen behind in terms of commercializing AI. Internally, the company has been rushing to release products to maintain AI supremacy.
TPU v4
On Tuesday, Google announced that it had built a system with over 4,000 TPUs joined by custom components designed to run and train AI models. The system, called TPU v4, has been operational since 2020 and was used to train Google's PaLM model, a competitor to OpenAI's GPT model, for over 50 days. According to Google researchers, the TPU v4 supercomputer is 1.2x-1.7x faster and uses 1.3x-1.9x less power than the Nvidia A100. They claim that the performance, scalability, and availability make TPU v4 supercomputers the workhorses of large language models.
TPU v4 vs. H100
It is important to note that Google's TPU results were not compared with Nvidia's most recent AI chip, the H100. The H100 is more recent and made with more advanced manufacturing technology. Results and rankings from an industry-wide AI chip test called MLPerf were released on Wednesday, with Nvidia CEO Jensen Huang stating that the results for the H100 were significantly faster than the previous generation. Huang wrote in a blog post, "Today's MLPerf 3.0 highlights Hopper delivering 4x more performance than A100. The next level of Generative AI requires new AI infrastructure to train Large Language Models with great energy-efficiency." It remains to be seen how Googles newest hardware will compare to the H100, as technically the H100 is much newer than the TPU v4, and Google likely has new innovations up their sleeve.
The Future
As AI continues to grow in importance, companies like Google and Nvidia are in fierce competition to develop the most efficient and powerful AI chips and systems. While Google claims its TPU v4 supercomputer outperforms Nvidia's A100, it remains to be seen how the TPU v4 compares with Nvidia's latest H100 chip. As the industry evolves, we can expect more advancements in AI chip technology, driving even more innovation.
The article:
https://www.cnbc.com/2023/04/05/google-reveals-its-newest-ai-supercomputer-claims-it-beats-nvidia-.html
Sources:
Add a comment
Tags: ML News
Iterate on AI agents and models faster. Try Weights & Biases today.