Prices

Apropos of yesterday's post on computer prices, it seems that back in 1982, James Fallows paid $4,000 for his computer featuring 64k of RAM plus another $800 for a floppy disk drive. According to the handy CPI calculator on the BLS website, $4,800 in 1982 is equivalent to a bit over $10,000 in today's money.

Naturally, I had no choice but to scroll over to the Apple Store and see how much computer I could get for $10,000. Well, I got myself a Mac Pro with two 3.0 GHz quad-core Intel Xeon processors, 8 gigs of RAM, four 750 gig hard drives, two Super Drives capable of reading and writing CDs and DVDs, a 30 inch Apple HD Cinema Display, a wireless keyboard and a wireless mouse. That comes to $9,449. I believe that's a 128,000-fold in RAM. The improvement in storage capacity is, in some ways, even more impressive.

Photo by Flickr user Rudolf Schuba used under a Creative Commons license

Comments

And...? Keep going, do the math to see if it fits Moore's law. 128,000 is about 2^17, and Moore says the doubling time should be 18 months (yes, that's for processing power, not RAM, but we'll use RAM.) 17 * 18 months is 25.5 years, which takes us back to January 1982. Pretty good fit.
Processing doesn't quite fit- 6 GHz / 128000 = 47 KHz. I don't think 8080 processors were ever that slow, I think the slowest ones were around 1MHz.

Posted by SP | July 13, 2007 2:01 PM

Try this:

I bought a 42 inch Sony Plasma HDTV in 2000. The list price was $16,000; as a Sony employee I got a 40% discount, so I got it for $10,000. Figure with inflation we're talking $18,000/11,500 roughly today.

It has gotten a lot of use and needs replacing. I am going to get a superior Panasonic model, same size, for under $1500.

Posted by hopelesspedant | July 13, 2007 2:05 PM

There used to be a rule that the computer you wanted cost $5,000. It held true for a long time.

I remember purchasing a Compaq386 for $8,000. I paid an additional $1,000 for 1 meg of RAM.

Think about that. You can now get gigs of ram for a few hundred dollars and back in 1988 I paid $1,000 to double the RAM in the computer to a whopping 2 meg.

Posted by neil wilson | July 13, 2007 2:25 PM

I think that system could definitely beat the WOPR.

Posted by frobisher | July 13, 2007 2:28 PM

Of course, every time computational capabilities (either in terms of memory or processor speed or both) double, the size of the problems you wanna solve on that computer doubles. And since not all algorithms are linear in terms of run-time or memory usage ... well, y'all get the picture.

Not to mention the fun of running programs written in spaghetti code ...

Posted by DAS | July 13, 2007 2:32 PM

Wow, 128000 times. That's quite a penis!

And what will you do with your magical penis?

Write essays, just like James Fallows did 20 years earlier.

Posted by jerry | July 13, 2007 2:36 PM

I looked it up on Wikipedia: to what I'm referring already has a name -- Wirth's Law.

Posted by DAS | July 13, 2007 2:39 PM

That's all the computer you got for $10K? Maybe coz you went to Harvard.

There's nothing special about the RAM you buy from Apple. Just buy the minimum 1G RAM from them and then go online somewhere and buy another 8GB there. it's incredibly easy to install. same deal with the hard drives. And the monitor ... well ... if you're willing to settle for a lowly 24" monitor, you can get one with similar specs as the Mac 23" monitor for $500. That'll leave you a *lot* of money left over for other goodies.

Posted by duh | July 13, 2007 2:40 PM

Not only that, but he'll write essays for the Atlantic. The more things change...

Posted by Obee | July 13, 2007 2:42 PM

A 128,000 fold increase in memory is roughly in line with Moore's law, which says computing capacity doubles every eighteen months.

It should have doubled almost 17 times since 1982 - and 2 to the power of 17 is 132,072.

Posted by Spike | July 13, 2007 2:54 PM

Ah, I see the first guy beat me to it...

Posted by Spike | July 13, 2007 2:55 PM

Sometimes I look at my flash drive -- 1 GB of storage, retail price $20 or whatever it was, a device smaller than a cigarette lighter -- and I think about the hard drive in my first PC (a 386). The hard drive was 40 MB and the size of a small brick; I bet it weighed five pounds. The whole PC was under $1000, but the hard drive probably represented a quarter of that cost.

Posted by JBJ | July 13, 2007 3:28 PM

Does this mean you'll accept the conservative critique of the CPI, and never cite the spurious real-wage stats that liberals like to think are representative of economic progress?

Posted by Chris | July 13, 2007 3:38 PM

64k is 2^16 bytes and 8G is 2^33 bytes, so it's actually exactly 2^17 times as much RAM.

You can't really compare CPU clocks directly. If you were going to, the dual quad Xeons are more like 8 processors than 2, so you'd want to compare 24GHz to the 2MHz 8080 in the SOL-20. That's only 12,000 times faster, but you have to consider that the each core in the Xeons has multiple parallel functional units, wider data paths, and much more cache (all of which mean it can do more with each clock cycle) I think Moore's Law is not violated.

Posted by Rhett | July 13, 2007 4:00 PM

Example #185638 for why inflation indicators have some serious problems.

Posted by Sebastian Holsclaw | July 13, 2007 4:10 PM

I loved the article. You are missing a key point: the computer wasn't bought in 1982. It was bought in 1979. He didn't buy the computer, play with it for a week, then bang out a review. He learned to program (spending 6+ months on tax software), he put in upgrades, his printer broke 10 times in 1 year (but his replacement broke 1 time in 1 year), and so forth. The 1982 article says his computer is old and pathetic and obsolete, and then goes on to list all the better computers - why would he buy an old pathetic and obsolete computer for so much money in 1982? He was lusting after computers far cheaper than $4000 in the 2nd page.

It's almost unimaginable today to use something for three years before writing about it, but so it is. In some ways this is more interesting than the computer specs!

And guys, you got Moore's Law wrong - it doesn't say anything about computing power (a near worthless term, all MhZs are not creating equal) or clock speed. It says that *the number of transistors in a IC will double every 18 months*, and that has held true mostly.

Posted by Joshua | July 13, 2007 4:12 PM

I think Professor Frink got this one right in "Much Apu About Nothing":

[in the late '70s]
[Frink stands in front of a huge mainframe]
Frink: Well, sure, the Frinkiac-7 looks impressive [to student] Don't touch it! [back to class] But I predict that within 100 years computers will be twice as powerful, 10,000 times larger, and so expensive that only the five richest kings in Europe will own them.
Apu: Could it be used for dating?
Frink: Well, technically, yes, but the computer matches would be so perfect as to eliminate the thrill of romantic conquest. Ha-ho-ha-hey-hoo.

Posted by Wrongshore | July 13, 2007 4:22 PM

Or you could build your own PC using Windows or Linux and get a more powerful, much more versatile machine for equal money. But I should know better than advocate drinking anything but the Mac Kool-Aid around here....

Posted by Freddie | July 13, 2007 4:37 PM

I did a comparison once between our home computer and a '50's machine, and of course the home computer was a gazillion times more powerful.
But the primary use for the old machine was designing the hydrogen bomb, and the primary use for the new one was was my daughter playing "Putt-Putt Goes to the Moon".

Posted by peter | July 13, 2007 4:56 PM

Or you could build your own PC using Windows or Linux and get a more powerful, much more versatile machine for equal money.

Oh, ffs.

The interesting issue is the way that the budget has changed. Fallows paid $800 to upgrade from cassette storage to a 5.25in floppy. Nowadays, if you're budgeting, it's smart to set aside as much in backup storage as you'd want installed: after all, it's one thing to have a knackered 100K floppy containing a single article, and another thing entirely to lose the contents of a 500Gb drive.

Posted by pseudonymous in nc | July 13, 2007 5:54 PM

Spike: A 128,000 fold increase in memory is roughly in line with Moore's law, which says computing capacity doubles every eighteen months.

It should have doubled almost 17 times since 1982 - and 2 to the power of 17 is 132,072.

(nitpick mode on)

It's 131,072, actually. And 8 gigabytes of RAM is not 128,000 times as much as 64 kilobytes. It is exactly 2^17 times as much. Remember, "kilo-"/"mega-"/"giga-" when used to refer to memory (as opposed to disk space) always refers to powers of 2; this is because of how the memory cells have to be laid out on the circuitry. 8 gigabytes of memory is 8 x 2^30 bytes, or 8589934592 bytes. And 64 kilobytes of memory was exactly 65536 (64 x 2^10) bytes.

It confuses everyone, I know.

Posted by Josh G. | July 13, 2007 7:12 PM

The interesting issue is the way that the budget has changed.

Another interesting perspective, I think, is that Fallows was paying for an insignificant-yet-probably-still-computable percentage of the total computer processing power available in the world in 1980. (Okay, this would be an easier case to make if we were talking about the 1960s or earlier, but still. . .) Nowadays whatever Fallows or any other individual is doing on a computer is completely lost in noise, because computation is everywhere. There must be some home PCs that do less, on average, than what goes on in a BMW 7-series engine.

Posted by RSA | July 13, 2007 7:13 PM

Ah, I see someone beat me to it.

Posted by Josh G. | July 13, 2007 7:13 PM

stay classy, man.

Posted by Freddie | July 13, 2007 9:21 PM

The guy in the next office has three monitors. I want to kill him.

Posted by leo | July 13, 2007 9:35 PM

Sebastian Horsclaw: "Example #185638 for why inflation indicators have some serious problems."

Maybe. Hedonic indicators of increased value would not track anything like 2^17, though. An order of magnitude increase in physical "power" is necessary for even modest qualitative improvement.

An order of magnitude increase (~10x) takes around 6 years*, which is not incidentally the approximate periodicity for major OS releases. [*For the pedants in the thread, going from 1 to 10 would be 5 years, but zero to 1 also takes time.]

In 1979, when Fallows first acquired his 8080 processor computer, with 48KB of memory, CP/M was king of an 8-bit world; by 1985, six years later, DOS was king of a 16-bit world; by 1991, Windows was inheriting a 16/32-bit world; six years after that, Win9x and WinNT were co-regents; by 2003, WinXP reigned, with MacOS and various Linux as rival but serene Princes.

Already, in 1982, when the article was written, Fallows had word-processing and could see spreadsheets and CompuServe on his horizon.

Word-processing became WYSIWYG during the course of the 1980's, but not much functional advance in word-processing has occurred since Word 2.0 for Windows was released in the early 1990's. If we were trying to deflate the nominal cost of word processing to account for technical advances since, say 1992, I doubt we could justify much of a tweak.

Posted by Bruce Wilder | July 13, 2007 9:59 PM

You guys are nerds.

Posted by bt | July 13, 2007 10:34 PM

[*For the pedants in the thread, going from 1 to 10 would be 5 years, but zero to 1 also takes time.]

This makes no sense. Going from 1 to 10 is increasing by an order of magnitude; going from 0 to 1, on the other hand, is increasing by infinitely many orders of magnitude.

Posted by micah | July 13, 2007 11:13 PM

Good point, micah; the "1 to 10" part seems to be about the base, the "zero to 1" part about the exponent. Also, for what it's worth, among the people I talk to (computer scientists) it would be completely unremarkable if someone used "increased by x orders of magnitude" to mean some value multiplied by 2^x. 10 is a common but not universal convention.

Posted by RSA | July 14, 2007 8:47 AM

My first actually usable computer cost me just under $3000 in 1988 - 12 MHz 80286, 40MB hard drive, 1024 KB RAM. I got as much work done on that as I do today on my 2GHz dual-core, 300GB hard drive, 1 GB RAM desktop at work today. It was a lot more fun too.

Posted by W. Kiernan | July 15, 2007 12:00 PM