Cont: Musk, SpaceX and future of Tesla II

[acbytesla]

He's saying it would be used to run the AIs, which would also be in orbit.

Not suggesting it all makes sense, but that bit does seem to.

I get his point/reasoning. AI computing uses a lot of power just to disapate the heat generated by the electronics. Which the cold space environment wouldn't require. And solar panels in space are 40 to 80 percent more efficient than on earth. Typically gallium arsenide. They also are typically around a whopping forty times more expensive per watt. A good 400 watt silicon panel can be purchased commercially for a $100. Whereas a GA panel of similar wattage is around $4000. Smaller and lighter makes them a no brainer in space because of the need for efficiency and the cost of putting it in space.

Now, no doubt increasing the scale of manufacturing GA panels would reduce their cost substantially. But still, there are many reasons that no one is mass producing GA solar.

 

[The Great Zaganza]

You are completely wrong.
Dissipating heat in space is almost impossible, because the vacuum of space is not near zero, it has no temperature and nothing to dissipate the heat to.
It's a major headache for plenty of satellite designs. The James Webb telescope is basically an incredible complex machine to dissipate heat with a telescope attached.

Solar panels in space are more efficient because a lot of the most energy rich bandwidth of sunlight is filtered out by the atmosphere, not because panels are cooled by space - in fact, overheating is absolutely a problem.

 

[The Great Zaganza]

But on top of that, bandwidth to send and receive data is way way worse than anything you would build on Earth.

Only someone trying to sell you their rocket services would tell you this is a good idea

 

[acbytesla]

You are completely wrong.
Dissipating heat in space is almost impossible, because the vacuum of space is not near zero, it has no temperature and nothing to dissipate the heat to.
It's a major headache for plenty of satellite designs. The James Webb telescope is basically an incredible complex machine to dissipate heat with a telescope attached.

Solar panels in space are more efficient because a lot of the most energy rich bandwidth of sunlight is filtered out by the atmosphere, not because panels are cooled by space - in fact, overheating is absolutely a problem.

I get that the problem of thermal management in space is a huge issue. Still, In the vacuum of space, the temperature difference between direct sun and shade is extreme, often exceeding 500 degrees. I also get that the lack of atmosphere prevents convection to dissipate heat. Now perhaps I misunderstand the problem and the benefits of solar and Ai in space. If not to keep the electronics cold then what's the benefit? There is no way in hell that they're going to putt a Terrawatt of solar in space.

 

[wea]

If not to keep the electronics cold then what's the benefit?

HDI High-Density Interconnect Investors

 

[The Great Zaganza]

AI is space:

Matrioshka brain - Wikipedia

en.wikipedia.org

 

[grmcdorman]

Cooling in orbit is doable, according to Scott Manley. Requires a bit of engineering, but it's a known and solved issue.

He comments, though, that they are apparently now talking about many smaller "data centres" that would be networked in orbit. Being focused on the aerospace engineering, he of course doesn't address the other issues that occur to me (at least bandwidth, latency and maintenance).

 

[arthwollipot]

You are completely wrong.
Dissipating heat in space is almost impossible, because the vacuum of space is not near zero, it has no temperature and nothing to dissipate the heat to.
It's a major headache for plenty of satellite designs. The James Webb telescope is basically an incredible complex machine to dissipate heat with a telescope attached.

Solar panels in space are more efficient because a lot of the most energy rich bandwidth of sunlight is filtered out by the atmosphere, not because panels are cooled by space - in fact, overheating is absolutely a problem.

Vacuum is in fact an insulator, which is why we use these:

Cooling in orbit is doable, according to Scott Manley. Requires a bit of engineering, but it's a known and solved issue.

It's possible, yes. The ISS does it. But it's not a slam dunk for putting hot things in space.

 

[novaphile]

People seem to forget that things get hot when in front of the unfiltered sun.

There's no convenient river/lake/sea to pollute with all that pesky heat.

 

[novaphile]

I'm reminded of people that don't understand why their moon pictures are always completely over-exposed.

I tell them: "It's an object that is lit by direct sun, set your aperture and and exposure for bright sun."

Them: "No no! That's not right, it's night time!"

(At that point I gracefully withdraw.)

 

[EHocking]

These would be the same people who think the phases of the moon are caused by the shadow of the earth.

What they understand a lunar eclipse to be is anyone’s guess.

 

[jeremyp]

I get that the problem of thermal management in space is a huge issue. Still, In the vacuum of space, the temperature difference between direct sun and shade is extreme, often exceeding 500 degrees. I also get that the lack of atmosphere prevents convection to dissipate heat. Now perhaps I misunderstand the problem and the benefits of solar and Ai in space. If not to keep the electronics cold then what's the benefit? There is no way in hell that they're going to putt a Terrawatt of solar in space.

I am confused. Why do you think there's a benefit?

Cooling in orbit is doable, according to Scott Manley. Requires a bit of engineering, but it's a known and solved issue.

It is solved. If it weren't solved, there would be no artificial satellites in orbit because the temperature would just keep going up and up until they broke. Nobody is saying it is not solved: they are saying it is very expensive. Data centres in space will never be competitive with data centres on Earth for various technical reasons.

The same applies to satellite internet, by the way. Starlink will eventually fail because it is only competitive in areas that do not yet have full fibre or 5G. The intersection of people who live in such areas with those who can afford Starlink is measurable in the tens of millions, which is not going to be enough to sustain the business, especially as that number is most likely to go down, not up.

 

[JayUtah]

Thermal design is solved in principle. However, solving it for a particular spacecraft design and mission profile remains challenging. Scott alludes to those challenges (which is all he can do in a 20-minute video) but the devil is always in the details. Showing that the budgets work out is the first step. If they don't, then no amount of engineering will get you there. But the budgets being within a margin of workable simply means you can now rationally attempt a design. In practice, thermal design has to proceed in concert with all the other design constraints coming from launch requirements, structural requirements, power budgets, pointing constraints, and so forth.

For example, Scott points out correctly that designing the heat flow immediately raises fluid flow issues if passive heat flow won't be enough for your solution. But he doesn't address that this is turn makes spacecraft dynamic control markedly more difficult. The shape and size of your piping not only has to satisfy the thermal design constraints, but also the spacecraft dynamic control in order to maintain attitude constraints. Moving fluids within your spacecraft has complex inertial effects. And when you realize that the efficiency of the radiator and the efficiency of communication and power generation are heavily tied to dynamic control, you see the complexity of the system you're trying to solve for.

Yes, these problems can be solved, but it's still a matter of competition with better and cheaper terrestrial solutions. You can put a Burger King on the ocean floor if you apply enough money and engineering. But why?

 

[mumblethrax]

If not to keep the electronics cold then what's the benefit?

It allows the companies creating enormous demand for data centers to pretend that they have a solution in mind for the capacity problems they're causing. It solves a PR problem, not an engineering problem.

 

[arthwollipot]

The problem is that an AI data centre generates a lot of heat, very quickly - much moreso than your average satellite. Can vacuum radiators cope with that level of heat generation?

 

[JayUtah]

The problem is that an AI data centre generates a lot of heat, very quickly - much moreso than your average satellite. Can vacuum radiators cope with that level of heat generation?

If they are large enough and if the heat flow from the CPUs to the radiators can be achieved. The problem is that the larger the radiator, the higher the infiltration pressure required to achieve permeation. If memory serves, it's proportional to the square of the radiator area, but I would have to go look that up to be certain.

 

[Andy_Ross]

Then of course the coolant pumps would generate heat

 

[JayUtah]

Then of course the coolant pumps would generate heat

...and consume electricity, and affect the handling dynamics of the spacecraft, etc. These are all engineering problems we know how to solve, but they have to be solved simultaneously for any given spacecraft design because their effects are tightly coupled. And in this case they are orders of magnitude harder than we've previously accomplished.

 

[Andy_Ross]

How about a parasol between the satellite and the sun?

 

[JayUtah]

How about a parasol between the satellite and the sun?

In Scott Manley's solution (the straightforward semi-magical one) the radiator would always be edge-on to the Sun and therefore have no view factor to it. However, it's practically impossible to have radiators on a spacecraft in Earth orbit that don't have a view factor either to the Sun or to the Earth, which is a substantial source of reflected solar radiation.

The problem with shading the radiator is that the shade might then block the view factor to deep space, which the radiator needs in order to be maximally efficient.

This happened on an Apollo mission. One of the LM's radiators ostensibly pointed upward and had a design view factor to open space. But the LM landed close enough to a mountain that the radiator could "see" part of the sunlit mountain. That amount of light falling on the radiator was enough to produce a measurable drop in radiative efficiency.