Follow @jack_ganssle

The logo for The Embedded Muse

For novel ideas about building embedded systems (both hardware and firmware), join the 27,000+ engineers who subscribe to The Embedded Muse, a free biweekly newsletter. The Muse has no hype and no vendor PR. It takes just a few seconds (just enter your email, which is shared with absolutely no one) to subscribe.

This month we're giving away the Zeroplus Logic Cube logic analyzer that I review later in this issue. This is a top-of-the line model that goes for $2149.



By Jack Ganssle

The Death of PCs

Published 9/21/2001

This past August a party in San Jose commemorated the 20th anniversary of the birth of the PC. Many of the industry's luminaries, from Gates to Grove, attended. HP's Carly Fiorina was there, as was her new partner Michale Capellas of Compaq. No doubt the event was a somber affair, with the influx of bad computer business news beating down stocks.

Reading of the bash I was left somewhat breathless, thinking of what has happened in this industry over so short a time. In 1981 the PC was just another offering in the fray of personal computers, which at the time was dominated by 8 bit 6502 or Z80 processors. Speculation was rife that staid IBM would never make a go of it in the topsy turvy world of personal computers, which at the time were all sold by very small companies.

Remember the specs? 4.77 MHz, floppy disks were options; the standard mass storage media IBM designed for was a cassette tape. 640K max memory, but no one could afford that at the time. No graphics. Basic was in ROM. I bought one of these, with two optional floppies, for $7000!

IBM's mighty marketing machine, clever bundling of the software, and the support of ultimately thousands of third party applications developers led to the demise of the non-IBM-compatible computer industry (except for UNIX players and Apple). The legions of small computer companies withered. It did take a few years for the PC to gain dominance. Even HP, soon to be the 500 pound gorilla of computer makers, straddled the battle lines, offering a model with both 8088 and Z-80 processors that ran DOS and CP/M.

The following twenty years taught us several "truisms". The Internet economy was not subject to profit and loss rules, needing neither products nor sales. There's never enough computer horsepower so the replacement market for desktops is infinite.

Someone asked Rockefeller how much money was enough. His "just a little more" answer mirrors users' attitudes to computer performance. But maybe a billion dollars is indeed enough; perhaps a billion Hertz and 50 billion storage bytes also satisfies virtually anyone's lust.

Traditionally Andy Grove giveth and Bill Gates taketh away. Rumor has it that Windows XP may indeed prolong this model, assuming anyone has much interest in a new version of Windows.

The real limits to performance are, I think, two: first a lack of desire or need for more speed. Just how much horsepower does a word processor really require? Isn't the speed of Internet apps more constrained by bandwidth than compute power? Sure, video editing, EDA tools, and a very few other applications do have a real need for speed. I'd wager that most computers are mostly idle, though, waiting for the user to do something - a fact exploited by the SETI folks who essentially networked millions of PCs to do signal processing when their owners weren't sucking up CPU cycles.

Second: a massive mismatch between processor performance and memory speeds. Your 1 GHz CPU executes a cycle in a nanosecond. How cool! But each non-cache memory access tosses something like 50 wait states into the execution stream. How bizarre!

Though cache attempts to bridge this gap the pathetically small L1 and L2 memories are defeated by the complex environments we use. At the moment I have 8 windows open, running applications consuming probably hundreds of megabytes. I bet the cache misses are off the chart. Yet the machine runs remarkably well, most applications responding with no noticeable delay.

Those forgoing Windows for Linux have even fewer performance woes.

So, as millions of disappointed upgraders have discovered, doubling processor speed results in but a marginal improvement in real application behavior. Despite record low PC prices, why spend a grand or more on something that offers so few benefits?

Thanks to this and the crumbling economy PC sales are down from last year for the first time ever.

Let's not forget that the PC was an offshoot of the embedded systems industry. A full decade of embedded microprocessor use passed between the invention of the 4004 and the PC's birth. Embedded processors outsell Pentium-class CPUs by a factor of 50.

Yet, when I go to a party and tell folks I'm in the computer business, they immediately think "PC", and ask for advice fixing their AOL setup or dealing with Excel macros. We embedded folks can't get no respect.