Memory and disk space will both expand to overflow existing technology. At the time Bill made his statement, few of us had ever used more than 64K bytes of memory. It was only after memory prices fell that the great explosion began.
I remember when 8K by 16 bits cost $1/bit. Compare that to today's costs and you can see why Bill assumed the growth of memory would be limited.
By the way, I have a 16K by 16 bit ferrite bead board framed in my living room as an example of the past example of memory technology.
In addition to the magnetic core implementation shown, there were other variations according to the number of wires through each core. There were 2, 3, 4, and 5 wire varieties. In the 2-wire type, all the bits in a row are accessed at the same time. The Control Data Star-100 used this type of memory.
Not to mention the absolutely wonderful "rope memory" used in the Apollo Guidance Computer systems. A form of almost physically indestructible ROM in which the memory contents were encoded (literally "hard-wired") by weaving (by hand) the sensor wires in and out of a string of ferrite cores.
Changing the contents of the memory meant unravelling the string and re-weaving the wires.
I remember these types of ROMs. I worked with one that worked as a character generator for a graphics display terminal, Computer Dispays Inc. ARDS.
It was made of wires passing, or not passing, through the magnetic circuits (ferrite posts). The bottom had a bed of posts and a lid that (I think) completed the magnetic circuit.
To edit, one had to lift the lid, clip a wire and then add a new one, going a different rout through the posts, connect the new wire and replace the lid.
This was really interesting. Good work. I'm dubious of one stat, though: it doesn't seem possible that the magnetic drum rotated at 750,000 rotations per second. In fact, that most definitely can't be true.
Sometimes in my HDL code when I use a delay line or Johnson ring I call it a mercury line. It's the same concept. The springs in old reverb units are similar--they're basically FIFOs.
I actually "touched" magnetic core memory in the late 80's in a professional capacity. It was used in some Allan Bradley PLCs that were in an industrial plant I worked in as a student engineer. Hard to believe it was still in use as little as 25 years ago in functioning equipment.
hmmm. in the mid 80ies there was a type of Intel memory that acted as SRAM, but upon power failure was able to write the whole array into EEPROM or Flash cells, in parallel, before power was gone. don't remember the name though...
Sperry Univac had a drum memory that was ~6 feet in diameter and ~18 inchs wide. Had a head per track so no moving parts except the drum. Because of the large diameter it did not have to rotate very fast to get a high speed at the R/W head
Besides programming computers that used core memory, I bought some from a computer museum along with a silicon wafer. I took both to work one day to show the other programmers. Our 17 year old genius programmer thought the wafer was pretty cool, but when he looked at the core memory, he turned to me and asked, "How does it hold a charge?"
Thank you Kristin! interesting article.
In the 1980s, for a few years, magnetic bubble memory was going to be the future of mass storage. Magnetic domains coded as North-South or South-North were steered around loops on a specially contructed ferromagnetic substrate.
Bubble memory was too slow and never quite made it and the Winchester disk (and its derivatives) only now becoming overtaken by 'Flash' solid state technology.
Magnetic memory, I was reminded, when recently clearing out my attic, that as a cross between magnetic bubble memory and the latest racetrack memory, the magnetic CrossTie memory was once expected to do great things in the world of disc replacement and NV memory. Its structure was characterized by a series of sawtooth shaped films facing each other with the magnetic bits stepping along between opposing teeth.
The other memory museum piece I found in the attic is a non-destructive readout magnetic square-loop core. This is a normal core with a second hole in the annulus that allowed the state of the main core to be read non-destructively. It appears to be wound with five wires. I think in application it was intended to serve as both a logic and memory device.
When I was programming conveyor systems I was introduced to, but thankfully never had to use, "Ball memory". It consisted of a series of wagon wheels(?) with a U channel carved in the rubber diameter. These wheels would be mounted a a single shaft with each wheel controlling a single diverter. The rotation of these wheels was geared to the movement of the conveyor belt. When a carton was destined for a certain lane, there was a solenoid that would inject a steel ball bearing into the U channel of the wheel. A microswitch would be positioned around the wheel at the location timed with the carton being at the diverter. The ball would be scooped out before making a complete rotation. A common problem was that the operators would forget to refill the input tray with the balls in the output tray.
I know, not really electronics, but the mercury delay memory reminded me of it.
Hmm. The reference to the 1401 and 36-bit words is wrong. The 1401 was a character machine, with each location consisting of 6 data bits, a word-mark bit and a parity bit. Based on the following reference to the 36-bit words of the 701, perhaps that was what was meant.
Yes, I programmed a 1401 in autocoder. It was my second computer and second assembly language. By the way, it is possible to see a live 1401 at the Computer History Museum in Mountain View if you are there at the right time. The smell of the mechanical card equipment sure brings back memories.