Yes they keep opting for efficiency in their development. That was the big leap when it came to deepseek, they were suddenly doing the same level of processing at a much higher efficiency. This trend persists in recent AI comparisons. The American models are more capable by several points, but far more expensive. It isnt going to be sustainable to live on that edge.
A while ago there was a report that looked at energy usage of data centers (in general, not just those for AI development/deployment). The energy consumption of Chinese data centers was only 25-33% that of the United States*.
*(the range is because it gives conflicting data for power in gigawatts and total energy in terawatt-hours)
I am absolutely certain without even checking that China has a fraction of the data centers that America has. So that doesn’t show anything to do with the efficiency of China’s data centers.
And after checking, yeah, China has ~10% of the data centers, so that actually points to them being much less efficient.
That can very well be because China simply bas fewer but bigger data centers and that it doesn’t need more, smaller ones because of a greater degree of economic planning and less “tech startups and a financialised economy looking for a money sink”.
Chinese computing might is much closer to the US’ in capability than Germany’s, which statistically has more data centers than it. And that’s with a high degree of tech sovereignty.
Yeah, you could just make anything you want up, sure. It “can very well” be because of anything, but the data you showed is literally useless for relating it to the comment you replied to.
the dams being constructed on the Yarlung Tsangpo are of the run-of-the-river type, in which water is diverted through tunnels to a lower elevation to generate electricity and then returned to the river. Such systems do not store significant volumes of water for extended periods, unlike large reservoir-based dams capable of extensive flow regulation.
Do you think they’d be able to be green and run models like this? Because that would be rad as hell.
Yes they keep opting for efficiency in their development. That was the big leap when it came to deepseek, they were suddenly doing the same level of processing at a much higher efficiency. This trend persists in recent AI comparisons. The American models are more capable by several points, but far more expensive. It isnt going to be sustainable to live on that edge.
A while ago there was a report that looked at energy usage of data centers (in general, not just those for AI development/deployment). The energy consumption of Chinese data centers was only 25-33% that of the United States*.
*(the range is because it gives conflicting data for power in gigawatts and total energy in terawatt-hours)
I am absolutely certain without even checking that China has a fraction of the data centers that America has. So that doesn’t show anything to do with the efficiency of China’s data centers.
And after checking, yeah, China has ~10% of the data centers, so that actually points to them being much less efficient.
That can very well be because China simply bas fewer but bigger data centers and that it doesn’t need more, smaller ones because of a greater degree of economic planning and less “tech startups and a financialised economy looking for a money sink”.
Chinese computing might is much closer to the US’ in capability than Germany’s, which statistically has more data centers than it. And that’s with a high degree of tech sovereignty.
Yeah, you could just make anything you want up, sure. It “can very well” be because of anything, but the data you showed is literally useless for relating it to the comment you replied to.
At least they’re trying. They just completed the worlds largest hydroelectric damn. So big people complained it slows earths rotation.
the upcoming “Nine Gorges” project doesn’t use a large reservoir which is nice
i couldn’t find anything on that is there a source?
https://en.natopedia.org/wiki/Medog_Hydropower_Station
nothing much, what’s medog with you?