Computing power rolled back 4 decades

[MadAxeMurderer 6,176]

Back in 1966 the Apollo guidance computer first flew.

A few years later it found the moon, and landed the lunar module.

 

The computer's performance was comparable to the first generation of home computers from the late 1970s, such as the Apple II, TRS-80, and Commodore PE

It clocked at 2MHz, had 2k of ram (15 bits wide), and 36k of rom (16 bit)

 

Of course NASA also had workstations, server/mainframes, and super computers

A workstation might have had 128k ram, 64k rom, and 256k on disk

A server 512k ram, and megabytes on disk

 

Now we can buy an Arduino for €20 with 2k ram, 32k flash, clock 12MHz

It's comparable to the landing computer

 

The pyboard is a work station with 128k ram, 512k flash

The esp32 a server with 520k ram 1M flash

And the rasberry pi for €40 is a super computer

[El Jeffo 42,106]

This thread would be ideal for the Singularity section. Where'd it go?

[MadAxeMurderer 6,176]

esp32 512k ram with wifi(server)

 

pyboard 128k ram with sdcard slot (work station)

 

Arduino uno with 2k ram (apollo landing module)

 

Mainframe or super computer

[yourkeau 7,719]

The fascinating thing is: people spread conspiracy theories that the moon landing was faked. They do not realize that at that time there were no computers powerful enough for such video and graph editing as we have today. It was technically easier to actually send a man to the moon than to fake it. 

[El Jeffo 42,106]

20 minutes ago, yourkeau said:

The fascinating thing is: people spread conspiracy theories that the moon landing was faked. They do not realize that at that time there were no computers powerful enough for such video and graph editing as we have today. It was technically easier to actually send a man to the moon than to fake it. 

 

bUt ThE fLaG tHeY pLaNtEd WaS wAvInG! SOUND STAGE!

[cb6dba 7,317]

We had computers 40 years ago?

Were they peddle powered? 

[MikeMelga 7,212]

1 hour ago, MadAxeMurderer said:

Back in 1966 the Apollo guidance computer first flew.

A few years later it found the moon, and landed the lunar module.

 

The computer's performance was comparable to the first generation of home computers from the late 1970s, such as the Apple II, TRS-80, and Commodore PE

It clocked at 2MHz, had 2k of ram (15 bits wide), and 36k of rom (16 bit)

 

The remarkable thing about that computer were power requirements, volume and especially the resiliance against radiation from Magnetic core memory!

Although this was the only way to do it in the 1960's. unfortunately this is still the current approach: radiation hardenend hardware. Then SpaceX came and introduced another concept: multiple computers who vote for a decision. Much more relisience, much more redundancy, much more performance and much cheaper.

[MadAxeMurderer 6,176]

Why would magnetic core memory not be resiliant to radiation.

 

I assume each bit was a charge pump. You get a current running through an inductor, that decays over say 10ms. So every 3ms you read it destroying it and crecreating it. A but like DRAM really. Would cosmic radiation or particles pack enough punch to flip an inductor of nano henrys?

 

Running current in an inductor is quite the energy store. That's how a charge pump regulator works. But the inductor is micro henrys, or even milli. But they could be pumping 10W

[yesterday 1,896]

31 minutes ago, MikeMelga said:

The remarkable thing about that computer were power requirements, volume and especially the resiliance against radiation from Magnetic core memory!

Although this was the only way to do it in the 1960's. unfortunately this is still the current approach: radiation hardenend hardware. Then SpaceX came and introduced another concept: multiple computers who vote for a decision. Much more relisience, much more redundancy, much more performance and much cheaper.

 

Military aeroplanes have had multiple computers voting to to take the correct action from the 1970's

 

There us nothing new on that

 

 

Almost all Airbus aeroplanes, use voting computers these days

 

 

[El Jeffo 42,106]

37 minutes ago, MikeMelga said:

Then SpaceX came and introduced another concept: multiple computers who vote for a decision.

 

I like you, Mike, but you really need to lay off the Musk Kool-Aid. Using multiple computers concurrently is a concept that was developed for the Space Shuttle.

[MikeMelga 7,212]

9 minutes ago, El Jeffo said:

 

I like you, Mike, but you really need to lay off the Musk Kool-Aid. Using multiple computers concurrently is a concept that was developed for the Space Shuttle.

I'm not stating SpaceX invented the concept, this has been used in multiple non-space projects before. But for rocket launches, everybody used radiation hardened computers, even when using voting systems.

SpaceX ditched the radiation hardened computers+memory, that was the main change, compensating it with voting system.

And it's a significant change. The huge increase in processing power means for example that trajectories can be recalculated in realtime.

 

Radiation hardened computers + memory are alwyas some 20 year behind, they require specialized developers and are really expensive.

 

40 minutes ago, MadAxeMurderer said:

Why would magnetic core memory not be resiliant to radiation.

They are. That's why they were used.

[Metall 14,483]

10 minutes ago, El Jeffo said:

Using multiple computers concurrently is a concept that was developed for the Space Shuttle.

 

I always thought airplanes had redundant autopilots before the space shuttle, but looks like I'm wrong:

https://www.nasa.gov/centers/johnson/pdf/584731main_Wings-ch4e-pgs242-255.pdf

 

[El Jeffo 42,106]

6 minutes ago, MikeMelga said:

I'm not stating SpaceX invented the concept

Sorry, but that's exactly what you implied:

 

52 minutes ago, MikeMelga said:

Then SpaceX came and introduced another concept:

[MikeMelga 7,212]

2 minutes ago, El Jeffo said:

Sorry, but that's exactly what you implied:

 

No, was talking about ANOTHER concept, which is ditching radiation hardened computers. Voting system is the way to ditch it.

[El Jeffo 42,106]

Not to beat a dead horse, but the Space Shuttle used a voting system, too.

[HH_Sailor 3,492]

So this is why my (grown up) kids still look in wonder if we happen to visit a computer museum.

 

I can name the bits and pieces, their uses and even say that I used them back when. 

 

They think I'm ancient. But in computer terms I suppose I am. 

 

The maths teacher at school in 1975 taught us Algol 60. Coding sheets sent with the post to Uni for processing and we got results a week later...

 

Kiel has a small museum, but Padeborn has the Nixdorf national museum.

Well worth a trip for those of an IT bent.

[MikeMelga 7,212]

4 minutes ago, El Jeffo said:

Not to beat a dead horse, but the Space Shuttle used a voting system, too.

Again, the point was to ditch the radiation hardened computers, which AFAIK, the shuttle did not, they were using something like hardened 80186 computers.

[MadAxeMurderer 6,176]

There are 2 aspects of radiation hardenning

 

A cosmic ray can flip a bit

Over time cosmic rays can destroy the gate physically, in the same way neutron bombardment weakens concrete in pebble bed reactors.

 

These are soft and hard errors. Voting can take care of soft errors but not hard errors

Back in the 90s my company actually sold EPROM emulators to NASA for their 32 bit mongoose radiation hardened cpu.

 

The engineer who made the purchase I met when he came to a conference near me.

He was quite the nerd who was delighted to explain radiation hardening to me.

 

Radiation hardened means it's immune to soft errors, and develops hard errors slowly.

Geometry is a particular enemy of radiation hardening, and they probably use 1000nm processes state of the art being 5-7nm

 

Larger geometries also make soft errors less likely. But soft errors cannot be designed out of hardware, even on land.

Giant server farms might have a say 1 in 1e15 probability of 1 bit being corrupted per second. But that's handled by the RAID or other disk master.

[Alexnf 2,263]

6 minutes ago, HH_Sailor said:

 

The maths teacher at school in 1975 taught us Algol 60. Coding sheets sent with the post to Uni for processing and we got results a week later...

 

 

I might be younger (44) but started programming quite early.

I remember finishing the University and thinking I had enough of "the ancient" stuff we've been taught, time to find a job doing something in a modern language.

First job I had was in Cobol (in banking) ...   

I still get 1-2 headhunter messages per month through LinkedIn/Xing about potential Cobol jobs...

[MadAxeMurderer 6,176]

OMG we got us a Cobol programmer.

When did you last code?

I wonder is there a Cobol compiler or vm for the rasberry pi? 

 

There's certainly such a thing in Linux.

 

I guess the numeric descriptor pic 999.999.99 was an essential part of financial programming where you wanted the system to trap on crazy or invalid numbers. 

Or was it just for formatting?