SoylentNews is people
SoylentNews
More details on Perseverance, at Gizmodo.
As we watched NASA put a rover on Mars last month, it definitely seemed like the agency had to be using some sort of high-tech processor in its machine. Surely the rover is built on something much more powerful than the components in devices us civilians use, right? But while NASA is technically using a specialized processor to power the Perseverance rover, it's not far removed from the world of consumer electronics—about 23 years ago.
NewScientist reports that the Perseverance rover is powered by a PowerPC 750 processor, which was used in Apple's original 1998 iMac G3—you remember, the iconic, colorful, see-through desktop. If the PowerPC name sounds familiar, it's probably because those are the RISC CPUs Apple used in its computers before switching to Intel. (Although now the company is back on the RISC train with its homegrown M1 processor.)
The PowerPC 750 was a single-core, 233MHz processor, and compared to the multi-core, 5.0GHz-plus frequencies modern consumer chips can achieve, 233MHz is incredibly slow. But the 750 was the first to incorporate dynamic branch prediction, which is still used in modern processors today. Basically, the CPU architecture is making an educated guess on what instructions the CPU is going to process as a way to improve efficiency. The more information that's processed, the better the chip gets at predicting what it needs to do next.
However, there's a major difference between the iMac's CPU and the one inside the Perseverance rover. BAE Systems manufactures the radiation-hardened version of the PowerPC 750, dubbed RAD750, which can withstand 200,000 to 1,000,000 Rads and temperatures between −55 and 125 degrees Celsius (-67 and 257 degrees Fahrenheit). Mars doesn't have the same type of atmosphere as Earth, which protects us from the the sun's rays, so one flash of sunlight and it's all over for the Mars rover before its adventure can begin. Each one costs more than $200,000, so some extra protection is necessary.
So, not off the shelf, exactly. One Million Rads? Nothing like a CPU that survives, even if you do not.
(Score: 2, Informative) by Anonymous Coward on Tuesday March 02 2021, @09:46AM (11 children)
PowerPC isn't dead, and military electronics are full of them.
(Score: 5, Insightful) by Rosco P. Coltrane on Tuesday March 02 2021, @10:46AM
For the military and aero industries, that's because they're certified. Getting a processor certified is a very long and very costly affair. Once it is, it's on the shelf for a very, very long time.
For the space industries, they have the same requirements, with 2 things added on top of that:
- The certs are even more stringent, because once it's gone, there is no replacement possible
- The electronics must be radiation-hardened, which poses its own set of problems
So yeah, space tech trails even further being aero.
(Score: 2, Touché) by gtomorrow on Tuesday March 02 2021, @12:18PM (7 children)
Easy, Captain Pedantic. Where exactly does it say PowerPC is dead in the "bullshit" headline?...or even the summary? Don't worry, I'll wait for you...
+2 Informative, my ass.
(Score: 3, Interesting) by SomeGuy on Tuesday March 02 2021, @02:04PM (6 children)
That is precisely what the clickbait headline is designed to imply by comparing a sophisticated space craft to a piece of hardware that consumertards are familiar with as something they have been told is "obsolete".
If they had used an oranges to oranges comparison to current embedded space or military hardware, or even newer Power chips used in IBM's current big iron or the Talos workstation, that would have been less of an implication. And fewer clicks.
(Score: 3, Insightful) by gtomorrow on Tuesday March 02 2021, @03:16PM (5 children)
No, that's what you (and the AC) read into it. I certainly didn't; I found it interesting -- not in a clickbait manner -- that 22-year-old tech was being used in the most exciting thing from NASA in a while.
Snob much? In any case, it is an Apples-to-Apples comparison (more than figuratively!). From the summary...
(Score: 1, Insightful) by Anonymous Coward on Tuesday March 02 2021, @06:28PM (2 children)
It's totally a click bait headline
The submitter could have simply said a "23 year old design radiation hardened PowerPC 750 processor" in the headline implying just how old the technology is (which ultimately is the story). Nerds would have understood what that meant. Liking it to a vintage apple computer would have made for good "in story" nod.
As it is, they might as well have photo shopped the lander with an "apple inside" logo. Now it's a story about apple technology in Mars rovers. Which isn't even true. Because that is the definition of click bait. You can read into it what you want, but clearly you missed the bait.
The headline was a complete success. And it's 100% click bait.
(Score: 2) by EvilSS on Tuesday March 02 2021, @06:46PM
(Score: 1) by hemocyanin on Tuesday March 02 2021, @10:11PM
I wonder if some element of the hardening is made more effective or easier due to the larger size of all stuff on the chip -- here on earth, we save power and get more speed by making chips every smaller, but as density increases, would it be easier for radiation to cause damage to the chip because everything is so close? Maybe -- I don't actually know but it seems at least superficially plausible and if that was the case, I'd expect this chip in Perseverance isn't the only thing "old" onboard.
As for "Apple Inside", it wasn't just used for Macs: https://en.wikipedia.org/wiki/PowerPC#Operating_systems_with_native_support [wikipedia.org]
(Score: 2) by sjames on Tuesday March 02 2021, @07:55PM (1 child)
But it's not actually 22 year old tech. The old iBook wasn't radiation hardened. Launch an iBook through the Van Allen belts and watch it fry.
(Score: 2) by kazzie on Wednesday March 03 2021, @02:29PM
</Peter Pan>
(Score: 2) by hendrikboom on Tuesday March 02 2021, @01:11PM
Not only not dead, but there's an open-source effort [libre-soc.org] to design a openPower processor.
(Score: 3, Interesting) by DannyB on Tuesday March 02 2021, @04:00PM
Back in the 1990s when I was still a card carrying Apple fanboy, I studied the Power PC instruction set and programming. I recall that I particularly liked the calling conventions. [informatimago.com] (this was back when everyone still had land lines for calling.)
The calling function calls a called function. Function Z might call functions A, B and C.
The calling function must have a fixed (known at compile time) pre allocated stack frame that has room for the parameter list of any functions it may call. Thus function Z must have a big enough parameter area within Z's stack frame for the longest parameter list of A, B or C.
The called function A, B or C, didn't have to allocate a stack frame unless it needed to. It didn't even need to save the return address register onto the stack unless it was going to call some other sub routine. In looking at RISC-V I noticed it had a register used for return addresses when branching. Thus a leaf function (which calls no deeper functions within the call tree) doesn't have to do anything with the stack. Merely return to the address that is already in the return address register. Leaving the stack pointer exactly where it was when the function was called.
I could see why Apple picked Power PC. Alas, Apple ended up switching to Intel because nobody made Power PC processors suitable for desktops. Motorola focused on smaller, slower, lower power applications like mobile phones, while IBM focused on bigger, higher end, fire-breathing server processors, that even the very lowest end chips were a bit too hot, power hungry and powerful for desktop applications.
Would a Dyson sphere [soylentnews.org] actually work?
(Score: 5, Insightful) by ledow on Tuesday March 02 2021, @11:03AM (12 children)
You *never* use state of the art stuff for anything serious in things like spaceflight.
You use well-tested, well-debugged, known-errata, millions of chip-years in real-world usage, hardened versions of previously-commodity chips.
For a start, the faster and more modern the processor, the more likely it's susceptible to radiation and mechanical disturbance (PCs on the ground tend not to move much, or operate in a radioactive environment) which could cause all kinds of problems.
Plus, it's simply not necessary. Perseverence's uplink to the Mars Orbiter is only 2Mbps. Collecting and processing data to that speed isn't beyond any chip made in the last 30 years. With room for real-time analysis, compression, encoding, etc. there's really no need to go mad and put some octo-core 5GHz thing there.
Not to mention heat, power, etc.
(Score: 4, Insightful) by driverless on Tuesday March 02 2021, @11:46AM (10 children)
Surprisingly, the opposite is true: The more modern the processer, at least for server and desktop, so x86, Sparc, and Power, the less susceptible it is to faults. If you take rad-hard as 500 kRad(Si), which is the limit set by ITAR for export-controlled technology, then a Pentium III from 2000 would take that. Ten years later generic AMD desktop CPUs showed no failures at 17 MRad(Si), over thirty times the ITAR limit, and no the 17 MRad figure isn't a typo. Similarly, IBM Power and Sun/Oracle(ugh) Sparc are tested with neutron or proton bombardment to fairly high radiation levels with no failures. For example Sparc has SECDED ECC on caches, parity checks on things like TLBs with failures regarded as a miss, mod-3 arithmetic checking on the ALU, etc. All the server CPU vendors claim a MTTU of decades for their devices.
So pretty much any desktop or server CPU from the last ten to twenty years or so is radiation-hardened, it's just not sold as such because having Taiwan fab your rad-hard export-controlled CPU is a bit of a problem.
(Score: 2, Interesting) by Muad'Dave on Tuesday March 02 2021, @12:16PM (3 children)
Citation, please?
(Score: 3, Informative) by driverless on Tuesday March 02 2021, @12:46PM (2 children)
Bunch of papers from space tech conferences over the last 20 years or so. Googling the 17 MRad result, it's here [nasa.gov], complete with the not-a-typo comment:
(Score: 2) by Muad'Dave on Tuesday March 02 2021, @02:13PM (1 child)
Thank you for the citation. That was a 32 nm processor - today's processors are closer to 5 nm.
Also, the 17 Mrad you mention is the total _dose_ - while that's a meaningful stat for long duration exposure, the dose _rate_ is at least as important (as satellites pass thru the van Allen belts, for instance). From the document - "Processor upsets occurred above 1.8 x10^9 rad(Si)/sec". That's a whopping dose, no doubt - 1,800 Mrad/s in fact. You total dose of 17Mrad took only 9.5 mS of exposure to that dose rate.
A satellite shielded by 3 mm of aluminium in an elliptic orbit (200 by 20,000 miles (320 by 32,190 km)) passing the radiation belts will receive about 2,500 rem (25 Sv) per year [wikipedia.org]. Assuming the rem to rad conversion factor of 1 for photons, that means the total dose per year is 2,500 rad - it would take a very long time to accumulate 17 Mrad from just the van Allen belts. That same article mentions that the "Astronauts' overall exposure was actually dominated by solar particles once outside Earth's magnetic field", so that needs to be taken into account as well.
For what it's worth, GPU's are very far behind CPUs [nasa.gov] in radiation hardness.
(Score: 2) by driverless on Wednesday March 03 2021, @12:17AM
Yup, and processors have been getting more and more rad-hard over the past 20 years or so, so I'd expect current ones to do even better. There was a comment at a conference some years ago saying that modern Intel CPUs are practically rad-hard components.
Sure, but that's the way it's evaluated. You typically only get a short window for your eval, in Europe at least it's usually done in facilities in former Soviet-bloc countries or Russia, and you can't afford to have it sitting there for months gathering data so you characterise it with a brief high dose.
(Score: 2) by Subsentient on Tuesday March 02 2021, @04:16PM (2 children)
Not only that, I haven't had noticeable ESD damage to any components since I was a child. It also seems that they've somehow become much more resistant to static electricity. I don't know why, but it seems to be an absolute truth, at least for me here in dry, static prone Arizona.
"It is no measure of health to be well adjusted to a profoundly sick society." -Jiddu Krishnamurti
(Score: 0) by Anonymous Coward on Tuesday March 02 2021, @04:44PM (1 child)
There is a simple explanation for this. While we might expect smaller feature sizes to lead to more ESD damage, what has happened is that essentially every integrated circuit made today has some sort of ESD protection circuitry built into it.
So yes, it follows that products made today with modern ESD protection are frequently more resistant to ESD damage than products of yesteryear.
ESD damage still can be a problem, however, especially when handling individual parts before they are mounted on the board. The protection is not perfect and ESD damage often results in very subtle failures that are hard to diagnose. It's still a good idea to take basic ESD precautions when handling electronics.
(Score: 2) by lentilla on Wednesday March 03 2021, @07:47AM
Ah, the beginnings of a cargo cult [wikipedia.org]! Not that I disagree with you - I always earth myself just in case.
Thank goodness for teenagers: they occasionally prove that doing dumb things quite often works (those things that the old folks regard as special and precious because they were, when they were young.) One of my favourite quotes (George Bernard Shaw):
So I suppose I can pen a new one: "It is a teenager's gift to demonstrate the old men's lore is outdated. It is an additional gift to give the old men a chuckle when that same teenager discovers that that the old ways are followed for good reason."
(Score: 0) by Anonymous Coward on Wednesday March 03 2021, @09:38AM (2 children)
(Score: 2) by driverless on Wednesday March 03 2021, @11:06AM (1 child)
That's why I'd take a commodity x86 CPU and way underclock it. Those things are actually quite energy-efficient if you're not pushing them for maximum performance 100% of the time.
(Score: 0) by Anonymous Coward on Thursday March 04 2021, @06:05AM
(Score: 2) by kazzie on Wednesday March 03 2021, @02:36PM
Except in the 1960s, when ICs were cutting edge tech, and they were used in Apollo's Guidance Computer [wikipedia.org] (though some thought it a controversial choice at the time).
(Score: 1, Funny) by Anonymous Coward on Tuesday March 02 2021, @11:47AM (1 child)
Mars Perseverance Rover CPU NOT vulnerable to Meltdown and all Spectre variants.
(Score: 3, Touché) by RamiK on Tuesday March 02 2021, @02:42PM
The summary clearly mentioned they have branch prediction so they speculate hence Meltdown / Spectre vulnerabilities...
compiling...
(Score: 3, Funny) by inertnet on Tuesday March 02 2021, @12:53PM
They could develop the next version based on a totally new concept: BI (Biological Intelligence). Cockroaches [fandom.com] can withstand the radiation but their intelligence still needs some improvement.
(Score: 0) by Anonymous Coward on Tuesday March 02 2021, @02:14PM (2 children)
I hear used iMacs are losing their resale value pretty quickly.
(Score: 0) by Anonymous Coward on Tuesday March 02 2021, @03:42PM (1 child)
So they are now losing their value like PCs always have?
(Score: 0) by Anonymous Coward on Tuesday March 02 2021, @05:50PM
Yes, Apple is finally catching up.
(Score: 0) by Anonymous Coward on Tuesday March 02 2021, @04:27PM
Esa Microprocessors [esa.int] are a bit outdated SPARC based, but they are radiation resistant.
(Score: 3, Interesting) by sjames on Tuesday March 02 2021, @08:04PM
In the embedded world, the idea is to meet requirements for computational power, not to maximize it. For example, while 32 bit ARM based microcontrollers are everywhere, 8 bit AVR microcontrollers are still common and in production. That's because if you underclock an AVR, you can just dump in power from an unregulated battery and it will run while the ARM may be a bit touchy about that. If the computational requirement fits within the capability of an underclocked AVR, it's still a good answer.
(Score: 2) by stormwyrm on Wednesday March 03 2021, @05:21AM
Numquam ponenda est pluralitas sine necessitate.