What a mouthful of a PDF lol. But as far as I understand, that PDF has nothing to do with datacenter cooling. Cooling a datacenter usually happens in a closed loop, meaning there is no place the water could evaporate (which is the closest thing we have to “consuming water”) to, so there is no loss. The water is cooled via a heat exchanger, which is not opening up the loop. We have the same concept with AIOs on PCs, and you don’t have to refill the water every now and then, because it doesn’t evaporate.
The PDF refers to power production (as most sources of power do rely on water), where there is certainly some amount of loss. But that is not what I was arguing against.
Ren’s most recent work focuses precisely on how AI is increasing water use. A large language model like OpenAI’s popular ChatGPT-3 must first be trained, a data and energy intensive process that can also boost water use. Ren found that training GPT-3 in Microsoft’s high-end data centers can directly evaporate 700,000 liters, or about 185,000 gallons, of water.
Once the AI model is in use, each inference, or response to queries, also requires energy and cooling, and that, too, is thirsty work. Ren and his colleagues estimate that GPT-3 needs to “drink” a 16-ounce bottle of water for roughly every 10-50 responses it makes, and when the model is fielding billions of queries, that adds up.
The researchers are saying otherwise. I tend to believe them
It’s very cool that you tend to believe them, but I’d like to understand how something in a closed loop is “evaporating” - that is physically impossible. I once heard they are planning to build datacenters in the ocean, but even then evaporation is unlikely as the datacenters won’t boil the ocean. The only way to make this work is if they submerge it in a small pond/lake or just flood the building, and keep dumping water into it - which is stupid aswell because there are MUCH better materials for that that are NOT conductive, like special oils, which are not water based.
So ye, believing researchers is one thing, but believing something that physically is not possible because it fits your narrative is stupid.
What a mouthful of a PDF lol. But as far as I understand, that PDF has nothing to do with datacenter cooling. Cooling a datacenter usually happens in a closed loop, meaning there is no place the water could evaporate (which is the closest thing we have to “consuming water”) to, so there is no loss. The water is cooled via a heat exchanger, which is not opening up the loop. We have the same concept with AIOs on PCs, and you don’t have to refill the water every now and then, because it doesn’t evaporate.
The PDF refers to power production (as most sources of power do rely on water), where there is certainly some amount of loss. But that is not what I was arguing against.
The researchers are saying otherwise. I tend to believe them
It’s very cool that you tend to believe them, but I’d like to understand how something in a closed loop is “evaporating” - that is physically impossible. I once heard they are planning to build datacenters in the ocean, but even then evaporation is unlikely as the datacenters won’t boil the ocean. The only way to make this work is if they submerge it in a small pond/lake or just flood the building, and keep dumping water into it - which is stupid aswell because there are MUCH better materials for that that are NOT conductive, like special oils, which are not water based.
So ye, believing researchers is one thing, but believing something that physically is not possible because it fits your narrative is stupid.
Evaporation, is my understanding. Even sealed containers have evaporation in heat conditions.