Dan Butler's
TNPC Newsletter

IRQ, IRQ, My Kingdom for an IRQ - Part 2

by Al Gordon

In the last chapter of this saga, I was discussing the difficulty of adding a SCSI card to a system without an available IRQ. I had every intention of making this follow-up a treatise on how to get around the problem.

One little problem: I couldn't.

Several readers sent me messages suggesting the time-honored remedies: free up an interrupt somewhere and assign the card to it. Alas, the time-honored remedies don't work as well as they used to. The wonderful irony is that plug-and-play technology has made defeating system defaults much harder than before.

The good news about plug-and-play is that you are now spared the tedious task of manually setting jumpers on peripheral boards to assign system resources. And 98 percent of the time this is unambiguously good news. The bad news is that in the 2 percent of the time you might need to make some kind of manual setting, most of the cards being produced today don't support it.

Traditionally, the way you got around IRQ limitations was that you disabled a device you didn't need -- typically an unused COM port -- and then assigned that IRQ to a card you wanted to add. But resource allocations in today's PCs are made by either plug- and-play operating systems such as Windows 98 or plug-and-play BIOSes. And many BIOSes provide little flexibility in assigning IRQs. (NOTE: technical support people I dealt with contend that the Phoenix BIOS used on Dell systems, such as mine, is especially inflexible, so it's entirely possible that I am seeing the ever-popular "worst case scenario.") Another irony is that what flexibility exists, typically is reserved for supporting old "legacy" (i.e., non-plug-and-play) cards.

An astute reader messaged me that PCI cards can share IRQs, so the lack of an IRQ should not be a problem. So I thought, also. Generally, that is correct. But some cards -- and the SCSI card I was trying to install is one of them -- require their own unique IRQ. Eventually, I got the card installed in a way that Windows 98 reported had no conflicts (although there still was a shared IRQ). But my system became totally unstable, which was nature's way of telling me: No Way. And while I had a COM port's IRQ to sacrifice, neither the BIOS nor Win98 would allow me to assign the card to it.

So, can anything be done?

There are several options. First, if it's a PCI card that doesn't require a unique IRQ, you don't have a problem to begin with. Second, as noted, you might have a less rigid plug-and-play BIOS, or some setup software for your card, that will enable you to assign an IRQ manually. Third, you might be able to find some cards out there with jumper overrides. ISA cards are still being produced, although they do not support data transmission rates as high as PCI cards.

Fourth, in the specific case of SCSI cards, if your system has USB (and in most cases, a system new enough to have this plug- and-play predicament will have USB), you might see if there is a USB version of the peripheral you want to attach. USB, incidentally, also opens up new possibilities for freeing IRQs. Your keyboard and mouse ports each consume an IRQ; switching to a USB keyboard or mouse frees up that IRQ. Companies, such as Belkin, that specialize in adapter doodads have released various printer, serial, and network port devices for USB, allowing you to assign some of those devices to USB.

(Please note: I happened to have had a run of good luck with USB, hence my bullishness about it. But as we always warn here at TNPC, your mileage may vary.)

And last, you can try playing musical chairs with your PCI cards. I found that swapping cards from slot to slot does change the IRQ assignments. I never came up with the configuration I wanted, but I suppose it is possible that if I had had the time to explore every possible permutation, I might have found it. However, unless you are very fortunate, don't expect to get much work done while you test this strategy.

But based on the last batch of reader responses, some of you have better ideas than I do. So here's a TNPC Reader's Challenge: given Windows 98, all plug-and-play cards, and a plug-and-play BIOS... suggest a strategy for forcing assignment of a PCI plug- and-play card to IRQ 4. No prizes, except the thanks of a grateful nation.

Meanwhile, the bottom line remains the same as it was in the first article: plug-and-play (and USB, for that matter) are patchwork fixes on top of outdated architecture.