Join 3,561 readers in helping fund MetaFilter (Hide)


Reinventing Magnetic Core Memory
May 12, 2011 7:37 AM   Subscribe

Magnetic core memory reborn is a project by Ben North and Oliver Nash implementing 32 bits of core memory using literal tiny core magnets on the Arduino board. The history and operation of core memory is explained and diagrammed. The Arduino has over 4,250 times this amount of memory standard.
posted by odinsdream (29 comments total) 18 users marked this as a favorite

 
Stringing core memory is definitely a skill that is no longer in demand. Until now, I guess.
posted by tommasz at 7:45 AM on May 12, 2011


I love this for the same reason I love redstone in minecraft. I love when computer circuits are blown up big enough that you can practically see them working -- it kind of makes the machine nature of computers much more clear.
posted by empath at 7:46 AM on May 12, 2011


This is an extremely good writeup.
posted by DU at 7:49 AM on May 12, 2011 [1 favorite]


This is an extremely good writeup.
Seconded. As I was reading, I was like "Okay, that project is really cool. But the writing! That's the major accomplishment of this project."
posted by Galaxor Nebulon at 7:51 AM on May 12, 2011


Anyone have a good estimate of the total amount of core memory produced? I wonder if it might be less than the amount of RAM in my current computer (though I suspect not).
posted by ryanrs at 8:05 AM on May 12, 2011


Highlight of the article...

A completely different approach to core memory, with a non-destructive read cycle, was invented by Eiichi Goto in the mid-1950s
posted by carlodio at 8:14 AM on May 12, 2011 [1 favorite]


Great post, the epitome of "found something cool on the web and want to share it?".
posted by Wolfdog at 8:15 AM on May 12, 2011


Interesting. lately I've been thinking how cool it would be to build an actual CPU out of relays or something. The 6502 had just about 3,000 transistors If you go to visual6502 you actually see them change as computation happens: all done in javascript. I imagine you might be able to build it with electromechanical devices (it would be a lot slower, of course). But how cool would it be to make a computer out of components that you can (in theory) make yourself?
posted by delmoi at 8:22 AM on May 12, 2011


A completely different approach to core memory, with a non-destructive read cycle, was invented by Eiichi Goto in the mid-1950s

So now you're telling me Goto is considered less harmful?
posted by benito.strauss at 8:23 AM on May 12, 2011 [5 favorites]


lately I've been thinking how cool it would be to build an actual CPU out of relays or something

Yeah that would be pretty cool.
posted by tylermoody at 8:24 AM on May 12, 2011 [3 favorites]


A favorite professor of mine in college, sort of an eccentric guy who taught a history of technology class (one that started, no joke, with how iron ore is mined), had a retired Data General Nova sitting in his office with one of the case panels off so students could see the boards. 8 or 16 kwords of core memory, I think. I always found it fascinating.

I was always tempted to try and get it working and I think I probably could have given enough time, but the disk packs for it had gotten lost and it would have required building a lot of interfaces. Never got around to it, although I'm certain it's probably still sitting there.
posted by Kadin2048 at 8:25 AM on May 12, 2011


... history of technology class (one that started, no joke, with how iron ore is mined) ...

This has always been the most fascinating aspect of technology, to me. Taking things back to their origins which is somehow both demystifying and humbling. I mean, go to 1:30 in this video where future Canon lens glass is being cooled off with a huge box fan. Glorious.
posted by odinsdream at 8:33 AM on May 12, 2011


I used to work on magnetic core arrays. Not the 'big' ones from mainframes (around 256K per stack), but small ones like this, around 1000-2000 cores, for buffers.
Under Calibrating the half-select current they talk about the current. We had a process that was called 'Shmooing the buffer', whereby you adjust the current to its optimal value. The arrays were sensitive to temperature and you had at least 2 different current variables. (see Wiki)
posted by MtDewd at 8:38 AM on May 12, 2011 [1 favorite]


"Core memory would probably have been a lot less popular had it been a write-only technology."

I'm not sure about this part. Perl is pretty popular, isn't it?
posted by FishBike at 8:53 AM on May 12, 2011 [4 favorites]


carlodio Took me longer than I want to admit to get that, because of pronunciation issues. I saw the first name and therefore read "Goto" as Romanized Japanese, which is pronounced "go toe". Apparently that was different enough from the English "go to" that it took me a few moments to realize the joke.
posted by sotonohito at 9:00 AM on May 12, 2011


As for core memory, Nthing the comment that seeing very large circuits is nifty and helpful in getting concepts across.
posted by sotonohito at 9:01 AM on May 12, 2011


The more things change the more they stay the same - the innovate S4 Super Simple Storage Service provides cloud based write-only storage at only $1/TB/month.
posted by ChrisHartley at 9:01 AM on May 12, 2011 [1 favorite]


Anyone have a good estimate of the total amount of core memory produced?

No GOOD estimate, but just based on the sheer volumes involved I'd estimate significantly less than a gigabyte's worth (with a 16 bit byte, BTW.)
posted by JB71 at 9:01 AM on May 12, 2011


That is really interesting and as others have said, extremely well written. I feel like I could build one of these after reading the report. Pop science publications should look more like this. (Are those little paragraph summaries on the left a common thing? Because they're great.)
posted by lucidium at 9:19 AM on May 12, 2011


Great article, thanks for writing. I'd like to see the same guys build some mercury delay line memory. It's a beautiful idea, storing bits as echoes in a reverberating medium.
posted by Nelson at 9:29 AM on May 12, 2011


A gigabyte would've cost around $100 million dollars. That seems too low for cumulative world production. I'd guess the true total is in the hundreds of gigabytes.

One datapoint: the CDC 6600 had on the order of a megabyte of core memory, depending on machine configuration. About a hundred were sold.
posted by ryanrs at 9:37 AM on May 12, 2011


delmoi: "Interesting. lately I've been thinking how cool it would be to build an actual CPU out of relays or something. The 6502 had just about 3,000 transistors If you go to visual6502 you actually see them change as computation happens: all done in javascript. I imagine you might be able to build it with electromechanical devices (it would be a lot slower, of course). But how cool would it be to make a computer out of components that you can (in theory) make yourself?"

Have you ever read "Code" by Charles Petzold? It's a really good ground up intro to the whole deal, and he basically discusses what you do. Not how to do it, but starting circuit logic with switches and how to build a latch and all that. Kind of over a lot of my head, but fascinating nonetheless.
posted by symbioid at 10:53 AM on May 12, 2011 [2 favorites]


Seconding "Code" by Petzold. It was sufficiently under my head (and my son's head) that I managed to build a one bit memory (self-link) and he learned how to make AND/OR/NOT gates which he builds in minecraft using redstone
posted by DU at 11:08 AM on May 12, 2011 [1 favorite]


Code is a fantastic book. It literally explains microprocessors by starting with telegraph switches.
posted by odinsdream at 11:14 AM on May 12, 2011


Delmoi, I want to say one word to you. Just one word.

Are you listening?

Snubbers.
posted by ryanrs at 11:21 AM on May 12, 2011


odinsdream, I'm watching the Youtube glass-making video you linked to. I recommend it to everyone. Jump to the 5:05 mark. It's like your usual automated production line film --- with FIRE!
posted by benito.strauss at 1:24 PM on May 12, 2011


benito.strauss; I'm pretty sure there was an FPP about those videos. If not, someone get on that!
posted by odinsdream at 3:11 PM on May 12, 2011


lately I've been thinking how cool it would be to build an actual CPU out of relays or something.

Oh my goodness you just made me nostlgic.

Back in the '80's, at WRPI, RPI's radio station, next to the Coke machine was a clock.

This clock was around the same size and shape as the Coke machine. It was driven by mechanical relays and made this pleasing hypnotic chunk-chunk-chunk-chunk noise, with an extra ka-chuck at the changes of minutes, and more at the changes of 10 minutes, and this triumphant ka-chunk-chunk-Chunk-CHUNK! at noon, fading back to the background chunk-chunk-chunk-chunk.

I can't believe it has so little web footprint -- just a couple of mentions here and there by old friends of mine. I wonder where it is now.
posted by Zed at 3:59 PM on May 12, 2011


I've been learning about digital logic lately. I also have an arduino. So wow, this is right up my alley; thanks.

I'm tempted to buy a whole bunch of ferrite rings now.
posted by wayland at 5:57 PM on May 12, 2011


« Older Why a mobile phone ring may make bees buzz off: In...  |  Osborne to target workers' rig... Newer »


This thread has been archived and is closed to new comments