And Not Even a Decent Funeral


  • Page 1 of 1
    Bookmark and Share

With the recent surge in interest in all things vampire, either through the book and hit movie Twilight or the HBO series True Blood, this emphasis on the dead who continue to stalk the earth seems like a fitting time to discuss our favorite technology vampire, the venerable ISA bus. Like vampires, ISA is dead–having been killed many times by the chip companies that drive our technology and the pundits that narrate their every move. Yet when the sun goes down, our ISA vampire rises from the grave and situates itself into embedded application after embedded application. Since ISA is already dead, they can’t kill it again. Only the proverbial stake to the heart could finally put ISA to rest for good, and there’s nary a stake to be found in the land of small form factor technology. Just a few feeble twigs like PCI Express.

About the only part of the analogy that falls apart is that ISA is toothless, let alone having fangs. Instead of being a ferocious bloodsucker, ISA is a slow, simple parallel bus that is ideal for the slow, simple serial, digital and analog I/O requirements pervasive in embedded applications. ISA is easy to implement, even for a second-year engineering student, requiring only hundreds of gates of a programmable logic device.

What was ISA’s sin that banished it to vampire-land? It uses a lot of pins–104 in some implementations. In today’s ultra-high-integration world, pins are an extremely valuable commodity. Silicon density has reached such levels that there’s almost no limit to the functionality that can be crammed into a modern processor or chipset. Silicon space is not the problem–having enough pins available for the I/O becomes the effective limitation. Even with the 1200-1600 balls available on some microBGA packages, it’s just too painful, not to mention archaic, for chip manufacturers to allocate about 100 of them to the ISA bus. So we have our ISA replacement–the LPC Bus–which, of course, you all know stands for Low Pin Count Bus. What a clever name!

LPC was created for the sole purpose of attaching devices that nobody in their wildest imagination would ever conceive of putting on PCI, let alone PCI Express. Devices such as a BIOS ROM, or Super I/O devices with serial, LPT, PS/2 keyboard and mouse interfaces–the x86 equivalent of a local bus. LPC was a convenient way to implement the critical parts of ISA to support these devices, while dumping a whole bunch of pins. LPC gets the pin count down from about 100 to 10. Not bad. Only a few ISA functions get lost in the process. 16-bit memory and I/O transfers. DMA transfers. Limited interrupt system. Things like that. The LPC Specification from Intel (who else?) says clearly that LPC was never meant as an ISA replacement or a board-to-board interface. But in the light of day, with the ISA vampire securely in its grave, LPC took on a life of its own as the ISA replacement.

Because, of course, one of the continuing strengths of ISA is the huge ecosystem of available I/O cards in a wide variety of form factors that has been built up over the last 27 years. And unfortunately, nobody told these suppliers that certain ISA functions would suddenly stop being available (along with a crippled address map, in spite of subtractive decode). So a new market opportunity arises–the chip that converts LPC back into ISA. The LPC-to-ISA bridge. Technology comes full circle, again.

Let’s recap. A simple, pervasive, ubiquitous technology is obsoleted prematurely so that the technologists can satisfy their oversized egos–the “if you build it, they will come” philosophy that has plagued the high-tech industry since its inception. Stop me if you’ve heard this one before.

I know, I know. We are beating a dead horse–er, vampire–er, bus. The ISA train has clearly left the station–or rather the train is still there but the station is gone. Maybe the next time our technology leaders want to invent something to fix the thing they invented to replace the thing that wasn’t broken in the first place, they will take a few minutes to contemplate and discuss what might happen if they don’t go forward–if that legacy capability continues to be available for untold dozens of years longer. A simple request.

Your comments and ideas are appreciated at