Are NASA computers really that powerful?

I've always seen jokes or memes comparing really good computers to NASA computers. However, I don't know whether their computers are really that good, and if so what computers do they use and why do they need to be so powerful?

NASA is a big administration spread across many centers, with many goals. This is reflected in the information we get back searching for what they use their various computing facilities for.

For example, The NASA Advanced Supercomputing Facility (lol, Advanced! Supercomputing, which will surely one day be eclipsed by a Very Advanced Supercomputing Facility) does a lot of fluid dynamics and climate simulation. That's at Ames in California, and it contains the Aitken system along with several other computing clusters.

(Sidenote: "supercomputer" has a fluid definition, and supercomputers of yesteryear are eclipsed by the computing power available in disposable electronics these days. These days a "supercomputer" also typically isn't a single computer, or even more suited to doing many common computational tasks; they're a lot of computers hooked together and their workload needs to be many nearly-independent tasks to really take advantage of their resources. One of the terms of art for such tasks is Embarrasingly Parallel.)

But NASA has a lot of software that just runs on normal consumer PCs, too. For example, the Copernicus trajectory optimizer, which is one of the tools used by the Artemis program, runs on regular Windows, Mac, and Linux laptops. The computers used on the International Space Station were upgraded to HP ZBook laptops in the past decade or so (and those were the laptops I targeted for the is-the-incoming-docking-Dragon-going-to-hit-the-space-station-or-not software, RPOP, for many years).

Maybe the best thing to say is: there's a lot of hyperbole about NASA. They're consumers of computer technology, not inventors. They buy high-performance computing from firms that offer it, and they do use that for some things, but the idea that there are supercomputers running every space thing because all of them absolutely have to work is nonsense. NASA computers are just like the computers of any other science/engineering institution.

NASA often has extremely old legacy computers being used. The US government, at least the NASA part of it, is not keen about spending money on replacing operating computer hardware until it becomes unavoidable.

The Shuttle Mission Simulator which was designed in the 1970s, used Univac mainframes until the late 1990s. The Univac terminal cluster controllers continued to be booted off of 8 inch floppy disks.

At some point after the late 1990s the simulation was ported to Silicon Graphics computers which were also obsolete. Part of the visual system ran on an Amiga personal computer. The Simulation Interface Device (SID), the interface between the (real) flight computers and the simulation used wire-wrap technology throughout the program.

This is the SID at the end of the shuttle program (personal photo).



In the Mission Control world, the Mission Operations Computer, which ran on a series of IBM mainframes, was used until 2002. Source: RIP, MOC

NASA employees and contractors also suffer with obsolescent office automation. Old versions of Windows running on old PC hardware is the norm.

Parts of NASA whose mandate require them to utilize cutting edge machines get to do so. But it is not the norm, and reading the hype about NASA computers in the media just put a wry grin on our faces.

Source: Lived it.

I think this is a carry-over from the 1960s when NASA really did need some of the most powerful computers in the world. This included both "big iron" - mainframes used to run mission control during missions and, just as importantly, to design equipment and plan missions in advance - and the small stuff. Those mainframes were far less powerful, at least in terms of CPU speed, RAM and online storage capacity (offline capacity effectively unlimited with tape libraries), than a modern desktop computer, and in fact compared to a desktop computer from a couple of decades ago. But they were among the largest computers at that time.

The small stuff, in particular, included the Apollo Guidance Computer. While the AGC isn't much by today's standards, being roughly comparable to a late 1970s 8-bit microcomputer, it was an amazing step in miniaturization.

While NASA's space missions continue to progress, the advances are far less in computer technology than in all the other engineering needed to get to space and work (whether people or machines) in space. Computer technology has improved dramatically since the 1960s, with Moore's Law as a good example of just how dramatically. But the number of spacecraft, missions of spacecraft, etc. (except perhaps for communications satellites, which for many years now have roughly paralleled increases in computer speed/capacity) have not increased at anywhere near that rate. In other words, NASA needed every bit of available computer capability (mainframe and micro) at one time but now, with some exceptions such as the already noted Aitken supercomputer, uses computers much the same as almost any other government agency or private company.

Are NASA computers really that powerful?

Yeah, they are. Very, very, very powerful. For example, aitken has a peak performance of 13.12 petaflops and was ranked number 58 most powerful in the world in 2022. For comparison, that would outperform your desktop computer in the range of 1,300 to 130,000 times!

Why do they need to be so powerful?

Because NASA is sending very expensive equipment or astronauts who would be expensive (both economically and PR-wise) to lose into a harsh and difficult environment they can’t survive in unaided where they can’t easily help them, and sending them atop a massive tower of explosives catapulting them into space at high speed, and maybe bringing them down at an extremely high speed and having them survive that as well.

That’s pretty complicated on its own, but now consider that they will need lots of renundacy as well. How would it go down if NASA said “sorry we lost $10 billion and 6 astronauts because we didn’t add a bit more power to the computer and it crashed, causing the life support to fail” or “sorry we didn’t realize the ISS was going to be hit by that asteroid and kill everyone because we didn’t have a computer powerful enough to simulate the influence of Jupiter on its trajectory.” Very, very poorly.

And, of course, it wouldn’t go down too well to say “sorry, our mission has been delayed by 10 years because our computers will take a while to run the simulations for its trajectory”. The computers need to simulate reasonably quickly.