• Please review our updated Terms and Rules here

Core Memory lifespan?

Z

Z80Jon

Member
Joined
Nov 20, 2016
Messages
15
I did a bit of poking around on the internet and wasn't able to find any specific answer, but does anyone know if there's any limitation to the number of times you can write to a given core memory location before it becomes unreliable? Hundreds of thousands of times? Millions of times? Billions?

I ask this question as I'm writing some software for a system with core memory that will be doing significant amounts of data swapping, and am slightly concerned that an excess of core memory write cycles could be unhealthy.
 
I don't think you can wear it out. It was used as RAM, so it was expected to be rewritten constantly. Also, I think reads are destructive, so any read requires a write cycle to put it back.
 
I read somewhere or heard that if a given single memory location is constantly and continuously accessed/modified for a long period (through a bespoke program) it could get warm but whether that was enough to damage the insulation coating on the wires through those ferrite rings, I don't know. Has anyone else heard this?
 
I don't think there is any realistic life, depending on the ceramic ferrite used.
Surely not from reading and writing.
Moisture can do some of type of cores in as well as eating the wires.
They say that they could be buried for 100K years and still be read
out as long as the temperature didn't go over the Currie point.
Dwight
 
1944GPW said:
I read somewhere or heard that if a given single memory location is constantly and continuously accessed/modified for a long period (through a bespoke program) it could get warm but whether that was enough to damage the insulation coating on the wires through those ferrite rings, I don't know. Has anyone else heard this?
Click to expand...

I can't see that this would be an issue. On most core based systems there are tight loops that continually access a single location. Some older IBM systems such as the 1620 actually run the core warm

http://www.angelfire.com/oh3/ebjoew/IBM_1620_Core_Memory.html
 
Thanks for the answers, guys. I suppose the supporting logic should be a bigger concern with respect to lifespan than the cores themselves, hehe.
 
The core itself will never fail unless physically damaged. The wiring however has been known to fail simply through oxidization caused by excessively poor operating or storage conditions.
 
The CDC 7600 incorporated a "duty cycle integrator" on SCM memory. The 7600 used very small core with a read/write cycle time of 275 nanoseconds, interleaved 10 ways, given a consecutive access rate of 27.5 nsec per 60 bit word.

A pathological programmer could figure out a way to keep hammering on a single memory bank, with the result that things would heat up and start throwing errors. The "duty cycle integrator" essentially detected too-frequent bank hits and added wait states to allow the core to cool sufficiently.

Good old memories. Consider that the 7600 dates from 1969. Darned fast for the time.
 
Thanks Chuck, so I wasn't dreaming it could happen.
You know, if you told us you could dip an unlabelled vintage magtape reel in ferrofluid, and - by eye, standing all the while - tell us whether it was data or source, and what version, somehow I would not be at all surprised :)

By the way I found the reference to core heating I alluded to above: http://www.psych.usyd.edu.au/pdp-11/core.html
 
You must log in or register to reply here.
Back
Top