We're in the I/O business here at EE Times: Every day, we sift through a boatload of hype and hidden agendas in the hope of telling true and fair stories. I'd like to share a recent example.
In March, I had an interview with an Intel manager who was surprisingly frank on a subject that I hadn't expected would come up: a broad Intel initiative on server I/O. He didn't reveal all, but he gave me enough information to whet my appetite.
A major wrinkle emerged a couple of months back via a Wall Street Journal scoop about PCI-X. That I/O enhancement, defined by Compaq, IBM and Hewlett-Packard, would not go head-to-head with the Intel initiative; it was a different class of technology aiming at a different time frame for deployment. But it was a shot across the bow.
More pieces of the story came together last month. It seems Intel has become the prime mover of the shift from shared-memory buses to CPU-independent switched fabrics. Its interest is understandable; the move enhances the power of PC servers and smooths the critical path for high-end processors-the only ones driving much profit margin for Intel these days. And all agree that the shift is inevitable.
But Intel is driving the transition so hard that it's riding roughshod over the concerns of its biggest customers, which have a storehouse of intellectual property in this area. The OEMs don't want to disrupt their ties with their increasingly cost-conscious end users by force-feeding a technology shift.
I'd call our story on this situation (see Nov. 30, page 1) a near miss. IBM wouldn't return my calls; I suspect the topic had become too sensitive. Hewlett-Packard would barely talk to me, and only on condition I wouldn't name the source. Compaq would go on the record but refused to answer most of my questions. As for Intel, I got a fairly good flow of information, but I still wonder whether it was the whole story.
We struggle with this kind of thing every day, so your help is invaluable. If you ever want to get honest-maybe painfully so-about a development in your neck of the woods, see the "How to Contact Us" box (in this issue, it's on page 152) and call the appropriate editor, or call me. We have plenty of smart editors. All they need is quality I/O.
