Nvidia says its GPUs are a 'generation ahead' of Google's AI chips
cnbc.comMy biggest impression over the past couple of days is that I’ve been working too long and have forgotten just how huge Google’s technological advantages are. From college until now, I’ve always relied on Google Search, even though I hardly use their other products. When I started working, it was Google’s open-source MapReduce paper that shaped the big data industry. Then came containerization, then Google’s Kubernetes. Nowadays, the entire cloud computing industry seems to be based on Kubernetes, and our entire team develops on top of it. Even this wave of Google’s stock price increase has benefited our company, because we have partnerships with them.
> Google Search
Unusable nowadays and monopoly that should be broken....
> Google’s open-source MapReduce paper that shaped the big data industry.
Experienced ETL practitioners laughed at that paper, knowing that was how DataStage and other Grid ETL platforms always worked. Imagine...spliting the data across nodes, working on it in parallel and then bringing it together...Oh my if was not for Google nobody would have thought about it...
> Then came containerization, then Google’s Kubernetes.
A platform full of security bugs, that they created because they lost the cloud wars:
Kubernetes: The Documentary [PART 1] - https://youtu.be/BE77h7dmoQU?t=122
They forgot about multitenancy, and 11 years after, they still dont have it....
> entire cloud computing industry seems to be based on Kubernetes, and our entire team develops on top of it.
Your team needs to go out more....
Nvidia GPUs have always been a generation ahead of the rest... Because no one else has seen any value is developing GRAPHICS processors for gamers.
Nvidia had an edge in CUDA support but they were and are not anywhere near a generation ahead in CUDA acceleration.
Take a step closer to present with tensor processing and Nvidia is actually just getting started. GOOGLE has more experience with datacenter clustered TPUs and tensor acceleration; they are just waiting for the h100 to reach its peak lifecycle (~4 months) before they obsolete it and "release" a catch phrase and model number.