NASA’s latest Mars rover has the same processor as an iMac from 1998

Krenum

Fully [H]
Joined
Apr 29, 2005
Messages
19,192
Lets hope the Martian PC Master Race doesn't find it. :alien:

https://www.theverge.com/tldr/2021/3/2/22309412/nasa-perseverance-mars-rover-processor-cpu-imac-1998

"Perseverance is running on none other than a PowerPC 750, a single-core, 233MHz processor with just 6 million transistors that’s most famous for powering the original “Bondi blue” iMac from 1998".

Link to Processor wiki:
https://en.wikipedia.org/wiki/PowerPC_7xx
1614727748157.png
 
Last edited:
I know there is probably a reason but I have to ask: Why? surely a cheaper and more efficient product exists than an old PowerPC on a micrometer process.
 
I know there is probably a reason but I have to ask: Why? surely a cheaper and more efficient product exists than an old PowerPC on a micrometer process.
The article explains it.

According to the article, its already had prior experience on Mars & it has a greater rate of survival than more intricut piece of hardware.
 
Last edited:
yes but they could have come up with something newer/better by now... 233mhz 256 ram and 2gb storage seems like so little for what its doing up there.
Hey, that's why they pay the people who make these things the big bucks right? To make the tough decisions :rolleyes:
 
Larger processes tend to be more tolerant of radiation and other abnormalities. On top of that, when you start the design and finally finish, it's YEARS before the thing flies. Also more likely to survive abuse like, oh, flying between the planets, since we're REALLY good at those older designs now.
 
Larger processes tend to be more tolerant of radiation and other abnormalities. On top of that, when you start the design and finally finish, it's YEARS before the thing flies. Also more likely to survive abuse like, oh, flying between the planets, since we're REALLY good at those older designs now.
yup. thats what the article said...
 
I know there is probably a reason but I have to ask: Why? surely a cheaper and more efficient product exists than an old PowerPC on a micrometer process.
I think both those variable would it be quite down the list of priority for a project like that.
 
Misleading article title, as the article itself states that the processor in question is a RAD750, a radiation-hardened processor released in 2001, based on (and fully compatible with) the PowerPC 750, and which costs upward of $200k.
Yup this is it right here. I remember back when I worked at NASA, and asked about a telescope I was working on at the time (now called the Spitzer Space Telescope) why they were using a 386 on it or something and not a pentium (this was circa 1997 or so) and got educated on how the pathways on the chip are just too close to be useful if a cosmic ray ripped through it could take out so many traces that the chip would be useless, and this was back in the day when they talked about the manufacturing process in microns not nanometers :D
 
Yup this is it right here. I remember back when I worked at NASA, and asked about a telescope I was working on at the time (now called the Spitzer Space Telescope) why they were using a 386 on it or something and not a pentium (this was circa 1997 or so) and got educated on how the pathways on the chip are just too close to be useful if a cosmic ray ripped through it could take out so many traces that the chip would be useless, and this was back in the day when they talked about the manufacturing process in microns not nanometers :D
Scroll down the wiki and you will see that they both use the same manufacturing process, they are the same processor. One just happens to be modified for radiation shielding.
 
Thing is that probe probably doesn't need more power than that. A lot of the control computers used in industry are pretty archaic by our standards here simply because better isn't necessary.
 
Thing is that probe probably doesn't need more power than that. A lot of the control computers used in industry are pretty archaic by our standards here simply because better isn't necessary.
Exactly - it’s not necessary to have more compute.

to an optimist, the glass is half full.
To the pessimist, the glass is half empty.
To an engineer, the glass is twice as large as it needs to be
 
Technically its not, its the same processor from 98, just an updated 2001 variant.
When one thing is $100, and another thing is $200 000, and there are good engineering reasons for choosing one or the other, I like to think that they are technically different. Technically they are about $199,900 different :p
 
When one thing is $100, and another thing is $200 000, and there are good engineering reasons for choosing one or the other, I like to think that they are technically different. Technically they are about $199,900 different :p
Radiation shielding and redundant pathways. You can break like 40% of one of those chips and it will still be completely functional.
 
NASA likes to use chips that have been around for a while. They favor consistency over speed and reliability over unknown bugs. A lot of hardware bugs actually take a while to find, and NASA needs to be able to trust these devices. Kernel panicks are extremely risky.
 
NASA likes to use chips that have been around for a while. They favor consistency over speed and reliability over unknown bugs. A lot of hardware bugs actually take a while to find, and NASA needs to be able to trust these devices. Kernel panicks are extremely risky.

Yeah. when you need to add triple-redundant voting logic to each gate on the chip, they would rather the rest of the chip be as simple and bug-free as possible.

Using an older process node also spreads-out gates further from each other / reduces the number of layers (less chance of a single strike taking out other layers / nearby gates with scatter radiation)
 
Last edited:
Its radiation hardened. Pretty much all that needs to be said.

I believe the copter is using a older snapdragon. The rover will probably live for years... even if everything goes perfect the chip running the coptor will be dead in months cause that snapdragon is going to FRY.

If you want to understand what radiation does to chips... just take any older digital camera and take a 2-3s exposure of a dark room. The image will be filled with noise... because individual pixels on the sensor will get hit and flipped. For a CPU to work in space where its taking 400x the radiation, or on Mars where it would be taking something around 50x the radiation as on earth. It needs to be thicker, with larger gaps between transistors to reduce the likely hood of large parts of the chip getting cooked by one high energy particle hitting the entire chip.

So you want a larger die... less dense transistors... and something we know how to make so well that you can etch it onto a triple thickness hunk of silicon PERFECTLY. As in the most cherry picked 100% operational chip you can imagine. There is a reason why the RAD750 costs 200k each... its on a triple thick hunk of silicon and its 100% operational AS in not one transistor isn't working perfectly. If they put a stock Ryzen or Intel CPU on a rover it would probably last a week or two tops. It will be interesting to see how long the ARM chip on the mars copter lasts.
 
Yeah. when you need to add triple-redundant voting logic to each gate on the chip, they would rather the rest of the chip be as simple and bug-free as possible.
Simplest thing might be to design with voting logic, rather than as an afterthought.

A three way vote with one vote jammed "1" gives OR
A three way vote with one vote jammed "0" gives AND
Straightforward three way voting also computes Carry.
A five way vote with two /Carries gives us a SUM.

A trio of nine way votes can three way vote with simultaneous triple redundancy.
No need to tack-on an extra time consuming step afterward.

Silicon on sapphire is usually preferred for rad hard.
Dunno why, since the silicon layer is still silicon.
For rads, Gallium Nitride might be the new shiznit.
 
Last edited:
yes but they could have come up with something newer/better by now... 233mhz 256 ram and 2gb storage seems like so little for what its doing up there.
Yes. You could stick an iPhone X up there and have 10x the processing power. You may also have to encase it in an inch thick of lead to protect it from radiation.

So what's the trade-off?

A CPU with 20+ years of experience and one everyone knows every single weakness, can optimize code for a thousand times over, but requires only a small sheet of aluminum to keep cool and shield from radiation?
-or-
A CPU that is insanely more powerful, smaller, less power sucking, but god forbid one stray ion cross it across 7+ months without an inch of obscenely heavy lead to protect it from deciding the Mars Rover is now the latest debris field on some random asteroid?

When my tax dollars are wasted on congress' golden toilets, war machines, and foreign aid, I'm glad that NASA takes the safe choice with what relatively small budget they do get. My hope in the future is that will no longer happen, but for now I will take what science advancements I can get.
 
Last edited:
Simplest thing might be to design with voting logic, rather than as an afterthought.

A three way vote with one vote jammed "1" gives OR
A three way vote with one vote jammed "0" gives AND
Straightforward three way voting also computes Carry.
A five way vote with two /Carries gives us a SUM.

A trio of nine way votes can three way vote with simultaneous triple redundancy.
No need to tack-on an extra time consuming step afterward.

But then you gotta hand-design and test the entire optimized chip, instead of starting with bug-free known working microprocessor, and just using proven-correct standard voting logic libraries replacements for each standard cell.

When you only start with a tiny 40 mm² chip, you have plenty of die space to waste!
Silicon on sapphire is usually preferred for rad hard.
Dunno why, since the silicon layer is still silicon.
For rads, Gallium Nitride might be the new shiznit.
Silicon on Insulator is the process IBM used to build these chips; it also reduces latch-up events. Sapphire might just be the chosen insulator because it's high temperature range
 
Last edited:
Larger processes tend to be more tolerant of radiation and other abnormalities. On top of that, when you start the design and finally finish, it's YEARS before the thing flies. Also more likely to survive abuse like, oh, flying between the planets, since we're REALLY good at those older designs now.
This. They are still building those old ceramic package 486 dx2? CPUs for satellite use for these reasons.
 
Yes. You could stick an iPhone X up there and have 10x the processing power. You may also have to encase it in an inch thick of lead to protect it from radiation.

So what's the trade-off?

A CPU with 20+ years of experience and one everyone knows every single weakness, can optimize code for a thousand times over, but requires only a small sheet of aluminum to keep cool and shield from radiation?
-or-
A CPU that is insanely more powerful, smaller, less power sucking, but god forbid one stray ion cross it across 7+ months without an inch of obscenely heavy lead to protect it from deciding the Mars Rover is now the latest debris field on some random asteroid?

When my tax dollars are wasted on congress' golden toilets, war machines, and foreign aid, I'm glad that NASA takes the safe choice with what relatively small budget they do get. My hope in the future is that will no longer happen, but for now I will take what science advancements I can get.

For related fun, it's kinda cool watching an unshielded digital camera get exposed to high radiation. The gamma photons kill individual pixels as they pass through the sensor. At moderate fields, it's a slow death. Looks like the typical snow that gets progressively worse over time. High fields it happens in seconds until the camera completely shits the bed.
 
The article explains it.

According to the article, its already had prior experience on Mars & it has a greater rate of survival than more intricut piece of hardware.
This ^ because the cpu MUST resist vibration, gforce, and most importantly cosmic radiation and solar radiation. There is a ton of stuff that is flying around on the particle level that would destroy, if not at least render a total lockup, of a modern cpu.

Its not about speed as much as its about stability. The code that nasa is putting through that little cpu is probably hyper effecient code to save time in transmission and processing. I think at the speed of radio in space it still takes about 6 mins to get to mars.
 
I know there is probably a reason but I have to ask: Why? surely a cheaper and more efficient product exists than an old PowerPC on a micrometer process.
left over game cube processors had to be used up? lol
 
This ^ because the cpu MUST resist vibration, gforce, and most importantly cosmic radiation and solar radiation. There is a ton of stuff that is flying around on the particle level that would destroy, if not at least render a total lockup, of a modern cpu.

Its not about speed as much as its about stability. The code that nasa is putting through that little cpu is probably hyper effecient code to save time in transmission and processing. I think at the speed of radio in space it still takes about 6 mins to get to mars.
11 minutes when the landing was taking place — that's an eternity when you have maneuvers seconds apart from each other. I'm sure the Perseverance team didn't mind using an iMac G3-era chip if it meant they'd have a working rover at the end of the day!
 
Hey, that's why they pay the people who make these things the big bucks right? To make the tough decisions :rolleyes:

You lack of understanding does not make you an expertise.
I really don't get why people are so quick to jump directly from "I don't understand why they did this" to "I am smarter than them"
You are literally using your lack of knowledge to say you know more...


Larger processes tend to be more tolerant of radiation and other abnormalities. On top of that, when you start the design and finally finish, it's YEARS before the thing flies. Also more likely to survive abuse like, oh, flying between the planets, since we're REALLY good at those older designs now.
Cheaper and efficient ≠ proven and reliable

now boys dont come here with understanding that his is not just a gammign computer you can reboot on a BSOD.Peope are trying to show they are smarter than the experts in the field here...
 
Like if we overclock this thing its going to drill rocks and sniff air faster? :) I like the explanation that Neil Degrasse Tyson gave about why the smaller you get the more macro physics has to give way to quantum physics, where you get smaller pathways for electrons to travel the less force it takes for something to affect it....even something that supposedly has zero mass, like Light Photons...when you get down into quantum observations, that was simple enough for my brain-meat to at least go "Hmmm, makes the sense" as drool slipped out the corner of my mouth and I debated Eating vs. Masturbating vs. Why Not Both........aka: You and Me ain't smart 'nuff to figure this out so just trust the big brains on it. They like a 486, we give them a 486. It won't run crysis, but it'll snif the air and take a month to move 300 yards.....but you know it's going to move the 300 yards.
 
Its radiation hardened. Pretty much all that needs to be said.

I believe the copter is using a older snapdragon. The rover will probably live for years... even if everything goes perfect the chip running the coptor will be dead in months cause that snapdragon is going to FRY.

If you want to understand what radiation does to chips... just take any older digital camera and take a 2-3s exposure of a dark room. The image will be filled with noise... because individual pixels on the sensor will get hit and flipped. For a CPU to work in space where its taking 400x the radiation, or on Mars where it would be taking something around 50x the radiation as on earth. It needs to be thicker, with larger gaps between transistors to reduce the likely hood of large parts of the chip getting cooked by one high energy particle hitting the entire chip.

So you want a larger die... less dense transistors... and something we know how to make so well that you can etch it onto a triple thickness hunk of silicon PERFECTLY. As in the most cherry picked 100% operational chip you can imagine. There is a reason why the RAD750 costs 200k each... its on a triple thick hunk of silicon and its 100% operational AS in not one transistor isn't working perfectly. If they put a stock Ryzen or Intel CPU on a rover it would probably last a week or two tops. It will be interesting to see how long the ARM chip on the mars copter lasts.
Yikes. What's it gonna be like for a human then, and neurons?
 
Yikes. What's it gonna be like for a human then, and neurons?
Well if you every watched Arnies Total Recall..... that first generation of colonists living under their cheap domes are all going to turn into mutants. Perhaps its a + many of us could probably use a wise voice to listen too even if its the voice of the man stuck in our chest.
 
Last edited:
Yikes. What's it gonna be like for a human then, and neurons?
Problematic. Hell, flying over the poles too many times can be problematic.

The other part - the RAD750 undergoes EXTREMELY long testing to verify that it's 100% and secure from radiation. They guarantee each chip for this kind of use. It's made by BAE, not Motorola (or NXP) anymore.
 
Back
Top