"It really seems like anyone with some renders and a white paper written by someone being gassed up by an overly agreeable AI can get VC funding these days."
We can’t fit tens of thousands of massive radiators floating around up there. The ISS has 27.2 meters / 88 feet of ammonia filled radiators that is able to expend… 56kW of heat.
Your 125 average is 2 and quarter times more that the ISS loop meaning we would need to launch up with each shitty satellite an additional 73 meters / 235 feet of trusses, support beams, gold and copper paneling and ammonia filled loops for each individual satellite.
Oh right and you want tens of thousands so at a reasonable 20,000 of those we need 1,500KM of radiators alone which is the radius of the fucking moon.
No, it won’t be that, cause its just as dumb and unfeasible.
The radiators are rated at 14kw a pair. There is other cooling mechanisms that are used and you are using the combined total. This scale and resource requirements become quickly ridiculous for launching that many radiator arrays alone.
Your math is all over the place. And this isnt any kind of scientific.
And raising the temperature factually increase the radiators capability to the 4th power, and the ISS runs at a cooler temperature than these GPUs need to run at, thus these radiators will be smaller. (edit: NASA source For a given amount of waste heat, the lower the rejection temperature, the larger the required radiating
surface. For radiation heat transfer, the quantity of heat rejected is proportional to the fourth power of
temperature; therefore, the area requirements increase rapidly as the temperature decreases. T)
I’ve been pretty consistent in this entire post on the 70kW but max rated for 84kW, but i might have been wrong about the 84 part mixing that up with something else.
I’ve been consistent with the AI datacenter being - 125kw avg, 150kw peak as well. 2x ISS before accounting for temperature difference is 140kW which is < 2x average, but not quite peak which is 2.15x.
There is no way these will be bigger than 2x what the ISS has.
Edit: Looks like you were working off of older data
the EATCS provides a substantial upgrade in heat rejection capacity from the 14kW capability of the Early External Active Thermal Control System (EEATCS).
We can’t fit tens of thousands of massive radiators floating around up there. The ISS has 27.2 meters / 88 feet of ammonia filled radiators that is able to expend… 56kW of heat.
Your 125 average is 2 and quarter times more that the ISS loop meaning we would need to launch up with each shitty satellite an additional 73 meters / 235 feet of trusses, support beams, gold and copper paneling and ammonia filled loops for each individual satellite.
Oh right and you want tens of thousands so at a reasonable 20,000 of those we need 1,500KM of radiators alone which is the radius of the fucking moon.
No, it won’t be that, cause its just as dumb and unfeasible.
The ISS is rated for 84kw and runs substantially cooler, they wont actually be that big. They will be smaller than the ISS whatever they end up being.
The temperature of the radiators scale to the 4th power.
It was more of a, it this is the worst case limit.
Edit: meant to say smaller than 2x which is the upper limit i said above.
The radiators are rated at 14kw a pair. There is other cooling mechanisms that are used and you are using the combined total. This scale and resource requirements become quickly ridiculous for launching that many radiator arrays alone.
Your math is all over the place. And this isnt any kind of scientific.
Straight from NASA
https://www.nasa.gov/wp-content/uploads/2021/02/473486main_iss_atcs_overview.pdf
And raising the temperature factually increase the radiators capability to the 4th power, and the ISS runs at a cooler temperature than these GPUs need to run at, thus these radiators will be smaller. (edit: NASA source For a given amount of waste heat, the lower the rejection temperature, the larger the required radiating surface. For radiation heat transfer, the quantity of heat rejected is proportional to the fourth power of temperature; therefore, the area requirements increase rapidly as the temperature decreases. T)
I’ve been pretty consistent in this entire post on the 70kW but max rated for 84kW, but i might have been wrong about the 84 part mixing that up with something else.
I’ve been consistent with the AI datacenter being - 125kw avg, 150kw peak as well. 2x ISS before accounting for temperature difference is 140kW which is < 2x average, but not quite peak which is 2.15x.
There is no way these will be bigger than 2x what the ISS has.
Edit: e.g post saying 84 & 70
https://lemmy.world/post/48895224/24572272
https://lemmy.world/post/48895224/24559267
Edit: Looks like you were working off of older data
Thank you for the update on the EATCS system.