Spacecraft processors are many, many generations behind the desktop. A state of the art Maxwell flight rated processor runs at 800 MHz, single core, no vector acceleration, and will cost you $500,000 to $1,000,000.
There is virtually no functional difference between a flight qualified processor and a ground development board. The cost difference is due to the overall reliability requirement for flight parts, for one. If you lose your spacecraft processor you lose your spacecraft, which probably cost you a couple of hundred million dollars. So the yield rate for spacecraft CPUs is quite low, like single percentage low. The chips that fail the qual inspection become ground development units.
In addition the manufacture, storage, and movement of each chip is very carefully managed and documented, which adds a surprising amount to the cost. This allows you to determine root causes in case of a failure, which then lets you understand whether other spacecraft using similar parts are in danger.
Radiation tolerance is an important part of this, both in terms of lifetime exposure tolerance and in terms of single event upset susceptibility.
Suddenly I understand the fuss about the early processors (eg, the Forth ones) that went into spacecraft in the 70s and 80s. It wasn't just the architecture that was good, it was the electrical/physical design that was noteworthy (which, to excuse the terrible pun, the CPUs are still writing home about :D).
Heh, I wonder how much the ground development boards cost? Probably way too much to find one affordably on eBay, say...
Also - how do you figure out if a processor is going to last for ages? How do you test it?!
Hmm, I get the impression some spacecraft probably have full failover capability as well.
And now I wonder what systems use CPUs in lockstep _and_ have full independent failover...
A ground development board would be on the order of $50,000, give or take. You'll also need a chassis and power supply of course.
“Also - how do you figure out if a processor is going to last for ages?”
Really big semiconductor junctions help. Low heat production also helps. And then there's this thing called “heritage”. If it worked on the last fifteen missions, odds are decent it'll work on the sixteenth.
Many spacecraft have fully redundant processors boards. Whether that helps you depends on whether you think chip failures are statistically uncorrelated. Which they are, sometimes. But not if there is a design flaw or manufacturing defect.
This is an absurd observation. Why is that surprising given that the satellite is in space and you have a giant powerplant connected to your PC and powering thousands of things in your city? More specifically, why is it saddening?
Probably because they think the satellite is more capable of doing cool things than their PC, which is just displaying HN and (presumably) nothing more for them.
No. It'll never leave the Solar System unless it happens to surpass the escape velocity of the Solar System. Then, it'll never leave the galaxy unless it surpasses the Milky Way's escape velocity. That would require a tremendous amount of energy. IIRC the Milky Way's orbital escape velocity is something between 300-500km/s (depending on whether you leave the Solar System in a way to take advantage of its forward velocity through the Milky Way or not..)
Sad to think my hacker news machine (PC) is using more power than an an intergalactic satellite.