The IBM 305 RAMAC is demonstrated at the Computer History Museum on Wednesdays at 1:00 p.m. It's an impressive machine to see in action. It has one arm for all the platters; the seek time is very slow as the arm is retracted, moves up or down to a new platter, and then engages. To keep the head floating above the surface, an air pump (maybe 1/4 HP) provides air flow.
The demo at the CHM is very impressive. They extracted the data from the drives many years ago and it turns out that it was a bunch of records about Canadian companies from the 60s or 70s.
> They extracted the data from the drives many years ago
There are a lot of drives in museums out there that might be worth reading. Same with the magnetic core memories (which are persistent and keep their data without power). For example, the only way that (maybe) someone could retrieve a copy of the flight software on the Saturn V flight control computer is if the Smithsonian ever lets someone crack open the unit in their collection and see if there's anything in that unit's core memory still.
Reminds me a bit of the fragile ancient scrolls scholars are now trying to scan with X-rays and make readable with new technology. Retrieval of the information is often balanced against a risk to the artifact's integrity.
Per Wikipedia the RAMAC 305 stored 5 million 8-bit characters[0]. Assuming they're using all 100 surfaces of the 50 disks in the unit that only comes out to 400,000 bits per surface. At the size of those platters the magnetic domains that encoded the bits must have been positively huge. There are products that could be used to visualize magnetic domains on tape[1]. The RAMAC platters seem like they'd be large enough that you could read them optically with one of these visualization tools.
The thing you'll see if you read the oral history, was they could have made it bigger, but 5MB was so much storage then, they were unsure how to sell a bigger one.
Somehow we manage to piss away computing resources like nothing, requiring more and more to do the same work. So I guess for meaningful comparisons we need a way to inflation adjust computing resources too?
E.g. 5MB of 1965 disk storage, roughly equivalent to 10 TB in the eyes of 2024 programs?
Maybe we could draw a comparisons by how much a base installation of a recent OS would eat away, or how much of the available RAM an at-the-time-modern text editor would gobble up? E.g. 1980s Emacs vs whatever currently popular Electron behemoth to compare RAM on computers from those respective epochs.
5MB of 1965 disk storage, as the article helpfully points out, stores 5 million (EBCDIC) characters of text.
5MB of 2024 storage, stores 5 million (ASCII) or as little as 1.25 million (UTF-8) characters of text.
There's no inflation to adjust here. There are still people programming microcontrollers with 1KiB or 512 bytes of RAM, 4KiB EEPROM, and maybe 1MB Flash. Those computers can do about the same things which 1965 computers with the same specs can do, just 20-100 times faster. Just because we don't use an AVR to add up Olympic scores, doesn't mean we couldn't.
The ones with five orders of magnitude more resources do a lot more. Some of it is squandered, when we can afford to do that, but in applications like AAA gaming or simulation, not so much. They operate near-optimally, just much faster, and doing a great deal more than was possible at the time.
Are we really "pissing it away"? For example the fact that VS Code is an Electron app written in HTML, CSS and TypeScript opened the door for many extension developers. I prefer it this way compared to the older native/Java-based IDEs.
Meanwhile, the reverse is true for me. I’m sick and tired of desktop chat clients and other applications being, or turning into, single-purpose Chromium instances.
The editor I'm using today updates itself from a global network of interconnected computers, using strong cryptography to ensure no one's tampered with it. It can download plugins from that same inter-net and has a sandbox that keeps malicious code from taking over the rest of its process space or accessing the filesystem. Some of those plugins make their own encrypted connections to distant computers. It displays its user interface on a truecolor screen across 2 2560x1440 monitors. It understands a huge number of programming languages and has advanced reformatting, error checking, syntax highlighting, search, building, testing, and reference-finding capabilities that can easily handle version controlled codebases with gigabytes of code. Because all its pointers are twice the size of those in older systems, I can open individual files many gigabytes in size.
Yeah, it's bigger than Emacs on a PDP-11. It does a whole lot more than Pico does, too.
At some point, 5MB of storage went from something that would kill you if dropped on your head, to something you might choke on if accidentally ingested.
(I'd put the cutover in the mid-1980s, when a handful of floppies would both be too large to swallow and too light to do much drop damage)
( 5120Kb / 360Kb = 14.2 floppies, and they weigh 18 grams a piece,
15 floppies at 18 grams is 0.27Kg = 9.5 oz.) Box and a half of 10.
ONE floppy would be too much to swallow, and although soft, its corner could do some damage if dropped on the sharp corner).
