NASA Notebooks Run Like x486 CPU

When I use to think of NASA I always thought the most powerful computers on the planet. I was surpised when I read this thread and the linked articles on a few fronts.

One being the performance of NASAs notebooks it uses on space flights. And two being the hand Intel played in NASAs supercomputing success.

NASA Space Notebooks Run Like X486

 

Very interesting. Seems like they would do most of the calculations down here on the ground...they probably only have monitoring software running on those computers in space.

That is something that most people don't think about - our atmosphere absorbs or deflects nearly all of the lethal radiation that those in space are subjected to.

 

Absolutely floored me, never would have thought they wouldn't have came up with a more effective solution for the radiation problem.

As you saw, the problem with doing most of the maths here is the transfer time to send then retrieve the data. They could get more accomplished if more could be done in space. At least till they get broadband for space connections

 

Having been through a sub-orbital system design in 2004/2005 I can tell you that radiation is a bear of a problem. It's hard to imagine that they can make survivable electronics for the Van Allen belt.

I'm quite certain that the shuttle laptops are running common lab software for space experiments; it would be ridilculous to take data from space to ground and get the results back - takes much longer than a slow laptop. In 2001 I was involved in evaluating some of the shuttle sub-system test stands ... and you'd be surprised at both how advanced some of the signal processing was in this 1970s pre-DSP era. You'd also be shocked at how low the total bandwidth was. The permanant comm links don't have the bandwidth of a cheesy old modem and they are not able to use that all for lab data reduction either. No one is surfing porn from the shuttle !

Some semiconductor technologies are much lesss susceptible to radiation single event upsets - (like Silicon-on-Insulator used on high-end PowerPC) but this just reduces the problem incidence, doesn't eliminate it. The problem is that memory(dram) technologies are not SOI and these generally are the largest radiation cross-section in the system. In fact the DRAM silicon technology details from Japan, Korea and the US will change overnight and the manufacturers don't notify anyone - it's impossible to radiation characterize radiation 'current' dram. You are basically stuck using common dram + ecc + special memory error scrubbing circuitry.

Triple (and more) redundant computing works reasonably well, and I strongly suspect thatthe currect software approach uses redundancy in software, not hardware. This has been demonstrated in several papers on space-borne computing - you run the same software in two or three different locations n memory and compare the results. This is pretty trivial to implement in applications. The OS is a different story - but read up on Xen and other virtual machine implementations and the future core vm level control on x86/AMD. [[sometime, maybe next year, you should be able to run Linux & Windows & whateverOS all at the same time on the same processor !]]. It does not secure the computation against all possible errors, but it's pretty good and pretty cheap.

One thing I don't understand ... all CMOS/NMOS circuitry (virtually every transistor in a modern laptop) is subject to radiation induced latch-up. A radiation hit causes a continuous current to flow through some gate and the only way to shut it off is to cycle power. *Maybe* a low power bit of circuitry like the CPU or the DRAM will last a while before localized heating burns out the chip, but for the power controllers chip you probably have milliseconds before the problem causes permanent damage. How do they handle this ? They probably don't - carry a spare since the laptop is not flight critical.

Hey - it's only rocket science - right ?
-Steve