Overestimate the gains and potential, underestimate the cost.
Typical corporate fads.
"We did it, Patrick! We made a technological breakthrough!"
A place for all those who loathe AI to discuss things, post articles, and ridicule the AI hype. Proud supporter of working people. And proud booer of SXSW 2024.
Overestimate the gains and potential, underestimate the cost.
Typical corporate fads.
Good thing we aren't running out of fresh water or anyth-
https://www.earth.com/news/satellites-reveal-a-sudden-drop-in-global-freshwater-levels/
Fuck.
"Running out of fresh water" narration is a complete lie. I don't know why, but these articles are implying its like oil that will run out some day. No, fresh water is a manufacturable good. Worst case scenario it gets more expensive as more of it needs to be produced out of salt water ect.
A big consumer in the fresh water market is agriculture. Whenever articles talk about demand exceeding freshwater supplies, it's referring to agriculture demand, which usually draw from dams, rivers, or lakes.
City water treatmant plants also usually start with pulling water from a river or water reservoir.
The costs with these consumers suddenly spinning up a saltwater or other advanced purification plant, that could perhaps function without a large freshwater reservoir, is prohibitively expensive. Especially for developing countries, where agriculture could be a large part of the economy
You do understand that the more expensive it gets, the less people can afford it, right? That's basically the same as running out. Sure, you can desalinate the whole ocean if you want to spend the resources on it. No one is suggesting that there will one day literally be no more fresh water within the foreseeable future. That doesn't really matter to the farmers in developing countries that can't afford to irrigate their fields and thus can't feed the people there.
And then there's the fact that desalination is a huge environmental disaster.
https://www.sciencenews.org/article/desalination-pours-more-toxic-brine-ocean-previously-thought
Where are you gonna get the energy to desalinate water from? What about areas that are thousands of kilometres from the ocean?
So whenever you see people using LLM-generated content, remember to ask them how much clean drinking water it cost.
It's not in a closed loop?
There tend to be 2 loops, an closed one to the server and an evaporative cooler as a cold side.
No, the water is evaporated as part of the cooling process.
What is this system called? I'd like to read up on it. Maybe there couldnbe regulations that those kind of systems need to use grey or recycle water. Maybe they already are in some data centers too
Well that's a stupid design.
That's physics. For a closed loop system, the cooling has to be done with air, which is less efficient and that doesn't work so well in say Texas in summer.
Making it a stupid design, yes.
That's how most data center cooling works, afaik.
I am so not an engineer, so maybe this is super stupid, but would there be some way to make it a closed loop by capturing the evaporated water and then letting it travel enough of a distance that it cools off and ~~liquefies~~ condenses and ends up in a holding pool?
Edit: I told you I was stupid.
It's actually how a liquid cooled PCs works. The warm liquid goes to a big radiator where fans blow air on the radiator to cool things. But lots of radiators becomes expensive and takes space. You're talking about a few hundred megawatt of heat.
The space would be a problem for sure, but couldn't you just use natural condensation to do it with a long enough pipe?
Yes, but you have to account for pressure differences. Steam condensing to water shrinks and causes big pressure changes. It's a lot easier to either vent it or use liquid everywhere.
Thats one option, part of strategies for reuse of liquid cooling.
To mention, its more energy efficient than air cooling, so there is a benefit. Smart companies though will also look to reuse strategies like using it for building heat. Larger companies will partner with the town/city to distribute the heat into town-wide systems, like for power generation or distributed heating systems, warm greenhouses, or even to dry out wood pellets for pellet stove systems.
Going long is effectively the same as using radiators though, you'll just need more pipe to do it without a radiator.
You're describing rain.
Unless you are talking about the entire planet, I'm fairly certain rain is not part of a closed loop cooling system for a server farm.
Probably expensive+ you're going to lose mass no matter what because physics don't give a shit.
It's not, if your only concern is lowering cost.
It improves the efficiency of the data centre because you need less expansive cooling. You need to get rid off the heat somehow.
"uses four times as much water as discoosed"