TE
TechEcho
Home24h TopNewestBestAskShowJobs
GitHubTwitter
Home

TechEcho

A tech news platform built with Next.js, providing global tech news and discussions.

GitHubTwitter

Home

HomeNewestBestAskShowJobs

Resources

HackerNews APIOriginal HackerNewsNext.js

© 2025 TechEcho. All rights reserved.

Powering cost-efficient AI inference at scale with Cloud TPU v5e on GKE

61 pointsby bobbypageover 1 year ago

6 comments

m101over 1 year ago
AFAIK Google is in the middle of massive infrastructure investment in hardware which is of similar performance to Nvidia, but they will own the whole stack. They will also be able to deploy an order of magnitude more compute than openAI to LLMs in 2024&#x2F;25.<p>LLM tech also seems to be easily swapped between providers. So my view is that the value proposition is in the hardware designers and manufacturers (Nvidia, tsmc, Google).<p>Why are more people not excited by what is happening at Google? What am I missing?
评论 #38454716 未加载
评论 #38455420 未加载
评论 #38456266 未加载
amrutha_over 1 year ago
Fantastic blog! In your post, you highlighted the use of Cloud TPU v5e with GKE for AI inference. How does this setup maintain high performance while managing costs, especially in high-demand scenarios like real-time data processing or live interactions?
评论 #38454530 未加载
nellydpa7over 1 year ago
The demo is really compelling. Using GKE or k8s sounds like a good idea for hosting LLMs, making them perform better and less costly overall. Not bad at all.
kernelsanderzover 1 year ago
a post about cost-efficient AI without any costs or comparisons?
hoangtrong001over 1 year ago
Great news
londons_exploreover 1 year ago
AI is a big deal this year. Google is one of the biggest tech companies.<p>Yet a post about AI at Google gets... No comments at all.<p>Back in 2005, this would have been the most talked about news of the day. Now, nobody cares what Googles up to. They lost their way.
评论 #38452696 未加载
评论 #38454693 未加载
评论 #38455246 未加载
评论 #38453865 未加载
评论 #38452910 未加载
评论 #38455162 未加载
评论 #38453569 未加载