Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Starcloud

blogs.nvidia.com

Post reply on HN
Print view View original

161–170 of 234 posts

Re: Starcloud

#161

Post by shrubble »

eptcyka wrote: post #105
preisschild wrote: post #78

Earlier quoted context omitted.

I wonder if "normal" RDIMM ECC would be enough to mitigate most of those radiation bit-flipping issues. If so it wouldn't really make a difference to earth-based servers since most enterprise servers use RDIMM ECC too

You'll get bitflips elsewhere besides just in RAM. A bitflip in L1 or L3 cache will be propagated to your DIMM and noone will be the wiser.

Sun Microsystems famously had this problem with their servers using the UltraSPARC II chips, with cache SRAM that didn’t have ECC. Later versions of their processors had ECC added.

Re: Starcloud

#162

Post by sketchysandwich »

Even if the some how solved the cooling problem others mentioned.

What happens when this data center becomes obsolete? we've just got a 4km wide piece of junk floating above earth now?

Re: Starcloud

#163

Post by glenneroo »

mickeyfrac wrote: post #159
burnte wrote: post #144

Earlier quoted context omitted.

Even beyond cooling, just getting all the hardware up there is extremely costly, and for what benefit over ground based DCs? The cooling is the ongoing problem but the cost of lifting it there obliterates all the other problems, IMO.

Space X thinks they will reduce the cost by 90% with Starship, so they are probably calculating off that.

On the linked page there are animations using Starship.

Re: Starcloud

#164

Post by righthand »

preisschild wrote: post #80
righthand wrote: post #72

Earlier quoted context omitted.

This used to be called fraud, now it’s cutesy lying?

Musk has been doing it for more than a decade now and didnt really face any real problems doing it...

Didn’t face any problems doing it… you mean when was charged by the SEC for lying on Twitter? Or do you mean when he was forced to buy Twitter to avoid another case against him?

Re: Starcloud

#165

Post by guluarte »

Reubend wrote: post #3

Last time these folks were mentioned on HN, there was a lot of skepticism that this is really possible to do. The issue is cooling: in space, you can't rely on convection or conduction to do passive cooling, so you can only radiate away heat. However, the radiator would need to be several kilometers big to provide enough cooling, and obviously launching such a large object into space would therefore eat up any cost s…

interesting, what if we put datacenters in the ocean floor with nuclear power? like the Army Janus program

Re: Starcloud

#166

Post by bbzylstra »

dabluecaboose wrote: post #147
LunaSea wrote: post #84

Earlier quoted context omitted.

Either the satellite is geostationary and doesn't have 24h / 24h sun exposure as energy source. Or they are not geostationary but it also means the datacenter will connect to a different earth base station which means the data access route would change and latency would increase which would be unacceptable for a lot of use cases. You would then need to replicate and synchronise customer data across the different spac…

> Either the satellite is geostationary and doesn't have 24h / 24h sun exposure as energy source. Due to the Earth's axial tilt [1], geostationary orbits generally have 24 hour sun exposure, except for a few minutes a day around the equinoxes [2]. [1] https://en.wikipedia.org/wiki/Axial_tilt [2] https://www.nesdis.noaa.gov/our-satellites/currently-flying/...

You can even see this in action via NOAA's CCOR-1: https://www.swpc.noaa.gov/news/earth-makes-appearances-goes-...

Re: Starcloud

#167

Post by btown »

notahacker wrote: post #52
Reubend wrote: post #3

Last time these folks were mentioned on HN, there was a lot of skepticism that this is really possible to do. The issue is cooling: in space, you can't rely on convection or conduction to do passive cooling, so you can only radiate away heat. However, the radiator would need to be several kilometers big to provide enough cooling, and obviously launching such a large object into space would therefore eat up any cost s…

You've also got the problem of cosmic radiation flipping bits. Your fault tolerant architecture probably mitigates this with redundancy, with the extra servers again eating into the purported advantages of extra solar power. Dealing with the PITA of single event upsets is something developers of edge data processing software in space put up with to avoid the latency issues that data clouds in space introduce

In all seriousness, if AI models can handle quantization, they can handle some flipped bits from time to time! There are probably some fascinating papers to be written around how to choose which layers in an LLM architecture could benefit more than others from redundant computation in a high-radiation environment.

Re: Starcloud

#168

Post by foobarian »

Reubend wrote: post #3

Last time these folks were mentioned on HN, there was a lot of skepticism that this is really possible to do. The issue is cooling: in space, you can't rely on convection or conduction to do passive cooling, so you can only radiate away heat. However, the radiator would need to be several kilometers big to provide enough cooling, and obviously launching such a large object into space would therefore eat up any cost s…

Hey, at least it's not going to end up with a bunch of actual people getting treatment based on invalid blood test results.

Re: Starcloud

#169

Post by torginus »

caminante wrote: post #150
yborg wrote: post #148

Earlier quoted context omitted.

This. We really have hit Internet bubble hype levels, this could be a Silicon Valley episode. "My pitch deck is one slide. 'It's AI... IN SPACE!'"

The banker slide with checkboxes will be compelling: [x] no permitting, cultural, wildlife [x] no local opposition. [x] site control Unfortunately, this is balance sheet financing so big boys only.

I can't wait for the inevitable epic humanity saving mission, where the AI datacenter gets stuck in a murder loop, and we have to send up the best and brightest in a spaceship to unplug the power cable and plug it in again.

Re: Starcloud

#170

Post by i_am_jl »

dabluecaboose wrote: post #154
Reubend wrote: post #3

Last time these folks were mentioned on HN, there was a lot of skepticism that this is really possible to do. The issue is cooling: in space, you can't rely on convection or conduction to do passive cooling, so you can only radiate away heat. However, the radiator would need to be several kilometers big to provide enough cooling, and obviously launching such a large object into space would therefore eat up any cost s…

Maybe this is reductive, but there are times that I'm concerned the only thing keeping me from getting gobs and gobs of startup funds are the facts that I understand basic principles of engineering in space. I could be wrong and this will be a slam dunk. To me, however, the costs/complexity (Cooling, SRP perturbation, stationkeeping, rendezvous, etc.) far outweigh the benefits of the Cheap as Free (tm) solar power

Assuming these people don't understand that their ideas are unworkable is a mistake. Don't believe for a second they are stupid or ignorant.

The difference between a criminal and a law-abiding citizen isn't that the citizen knows that crimes are wrong, it's that the citizen cares that crimes are wrong and the criminal doesn't.

Post reply on HN