Hacker News Reader

I have a 10kW 4090 training cluster in my garage with water cooling.

It paid itself back in 3 months if I'd used AWS services.

Not something that requires HA, backups, scaling, auditing, logs, security, and such, though.

Would you run your prod inference cluster the same way?

It's little brother is the prod inference cluster. It's had an uptime of 100% over the last two years.

People seriously don't realize just how _big_ computers have gotten. The AWS mentality is still stuck in 2005 when a t2.micro was a decent computer.

> The AWS mentality is still stuck in 2005 when a t2.micro was a decent computer.

The T2 EC2 instance family got announced in 2014: https://aws.amazon.com/blogs/startups/announcing-amazon-ec2-...

The first instance was the m1.small it's specs are marginally better than those of t2.micro, it's not been generally available since 2022 if memory serves.

https://aws.amazon.com/blogs/aws/ec2-instance-history/

https://instances.vantage.sh/aws/ec2/t2.micro

https://instances.vantage.sh/aws/ec2/m1.small

Please nitpick better. Nothing is more embarrassing then being pedantic and wrong.

And t2.micro was not a decent computer at that time.

You don't need most of that until you're making enough money to hire someone to worry about it.

is that enough to heat your house in the winter?

Reply on news.ycombinator.com