• Please review our updated Terms and Rules here

Core memory and the arduino

As much as I hate to hi-jack a thread, this post reminded me that I'm supposed to be looking for some hand-sewn core memory for a client for his display of "stuff". Anyone have some or know where some may be acquired?

Thanks
 
Druid6900 said:
As much as I hate to hi-jack a thread, this post reminded me that I'm supposed to be looking for some hand-sewn core memory for a client for his display of "stuff". Anyone have some or know where some may be acquired?

Thanks
Click to expand...
How can you tell it's hand-sewn? Weren't they all, or at least most of them?

I've got a couple; talk to me off-list, time for a get-together anyway.

I've also got some cores, in case someone wants to build one...
 
Last edited:
Never had the chance to see a stack of core with my own eyes let alone own a stack. I'm hoping I can eventually find something for a PDP or something.
 
Very pricey still. Always wanted one but never found a good one for under $100 :-( Haven't had time to read the article but have it bookmarked and look forward to it. I've always found that technology mentally interesting. Pretty strange concept since I grew up pretty much with RAM already being the standard although that probably would turn into a another very interesting article about RAM vs destructive memory. One of the more interesting books I picked up years back is called The Computer Science Encyclopedia (well crap I wanted to verify that title but can't find it so I could be mistaken).. anyway it's an early (first?) edition with a black cover dress but I think the book itself is grey if it doesn't have that cover. Either way great read and also has descriptions of early technologies that were being discussed and implemented.
 
g4ugm said:
Nice thing aboutcore is being magnetic its also non-volatile. switch the power off and on and it should still be there...
Click to expand...

I spent a good part of my professional life dealing with core and I don't miss it. Power-hungry, slow and problem-prone. If a manufacturer had to have any speed at all, interleaving of banks was mandatory; if you used SIMD instructions, not only interleaving was important, but also physical word width (I developed for a system that used 544-bit wide physical words).

Even so, you could run into troubles. Keeping things cool was an issue.

The CDC 7600 used very fast core that could overheat if the same address was accessed frequently enough. Seymour Cray had to incorporate a circuit that integrated the access rate over time and delayed accesses if core got too warm.

Similarly, the IBM 7090 kept its core in an oil bath, which made servicing of the core stack "interesting". I vividly recall a CE tool consisting of a broomstick with a magnet on the end to fish out bits of broken core from the soup.

Back in the early days of Kilobaud or Microsystems, I believe there was an article describing how a surplus 4K core stack was incorporated into an S100 box.

I think I still have a 1401 core frame kicking around in a desk drawer somewhere. When IBM started scrapping the 1401s, they were very common in surplus shops.
 
Consider the speed of the 1620--and then compare it to the 7090/7094. The 1620 used what, 12-bit core? (two digits per access); an instruction was 12 digits and the time for a multiply on the CADET was in the millisecond range. The duty cycle of the core wasn't high enough to matter.

I think of the 1620 and the 1130 mostly as "mini" computers; certainly the 1710 and 1800 variants used as process control boxes followed that line of thought. The big number-crunchers like the 7090 were very different animals.
 
tingo said:
That was my thought also; surely, all core memories was assembled (threaded) by hand?
Click to expand...

No, bulk core was often machine-made. I know that CDC ECS was so manufactured. Slower, but wider physical word, so transfer was still at mainframe speed once you got a block transfer started.
 
Chuck(G) said:
No, bulk core was often machine-made. I know that CDC ECS was so manufactured. Slower, but wider physical word, so transfer was still at mainframe speed once you got a block transfer started.
Click to expand...
Interesting. Are there any info about bulk core online anywhere? I tried a bit of googling, but that didn't return enough info for me to understand how bulk core would differ from "regular" core memory.
 
Not a lot of information, just a few ads and brochures, mostly.

In a lot of applications, bulk core was used to replace drum as a higher-speed memory, particularly for swap and paging.

In my particular project at CDC, we connected 4 Cyber 70s to 4M words (60 bit) of ECS and used it as swap space (we organized tasks into "chains" consisting of "modules", with inter-process communication between modules of the same chain, as well as between the lead (supervisory) module of each chain). We swapped modules into any available system for execution. Because there was inter-module communication, the OS kernel had to handle data movement between systems as well as programs--we implemented much of the COBOL move and arithmetic in the OS kernel. It was wonderful to watch when it was working right. It would have been far too slow with the disk drives of the day.

Google "bulk core memory" for some examples.
 
You must log in or register to reply here.
Back
Top