Some called 2001 the worst year in history. Others characterized it as the year that switch fabrics came back into the spotlight. Still others looked at it as the final emergence of the WLAN. But for me, 2001 looked like the year of the interconnect.
From as far back as the Gigabit Ethernet Conference in March, a large amount of discussions in the communication design sector began to focus on the bottlenecks created by the interface between the network processor and host processor in most networking designs. And as a response, companies like Intel, AMD, and Motorola began touting the wonders of new wiz bang technologies like HyperTransport, RapidIO, Arapahoe (3GIO), and more.
While all these technologies get pushed, what gets lost is wide scale adoption. Say what you want about PCI, but it has one very important thing going for it: The standard is well known and has become a defacto industry specification for equipment designs.
Unfortunately, the same cannot be said about things like RapidIO, Ara-pahoe, and the other new processor interconnects. The processor camps are clearly divided on what I/O technologies they plan to support. The MIPS crew, including Broadcom and PMC-Sierra, have put their backing behind HyperTransport. The Power-PC camp is going the way of Rapid-IO. And let's not forget Arapahoe. With Intel backing it, don't be shocked to see net processors, XScale processors, and Pentiums supporting this interface.
The problem with all these brewing battles is they are not necessarily the best for the design community. If designers wants to choose best-in-class parts, they could easily find a situation where one chip has a HyperTransport interface, while the others have a RapidIO interconnect. That means another piece of silicon, probably an FPGA or PLD, is needed to interface the processors together.
So, what can be done? A single standard is the holy grail. But, with the technologies so new, and the battle lines drawn, that's unlikely to happen. Another intriguing option is for processor manufacturers to deliver more flexible I/Os that support any of the above interconnects. Again, this is tough to build and, thus, a wish list item at best.
Of course, designers can base their component decisions on the processor interconnect. But then aren't you giving too much weight to the I/O?
The truth is there is no good option here. And that just means more headaches for today's equipment designers. So watch for I/O technology to remain as one of the hot topics throughout 2002.
---
Saying Goodbye
This marks my last issue as the Editor-in-Chief of Communication Systems Design. Starting next month, Patrick Mannion and Loring Wirbel will be heading up the editorial efforts on this publication.
But, I won't be going far. I'll be headed back to the web to head up the editorial team for Communication Systems Design's sister sites, www.CSDmag.com and www.CommsDesign.com. So, please stop by the sites and say hi. And keep those e-mails coming. rkeenan@cmp.com.
http://www.csdmag.com
© 2002 CMP Media LLC.
1/1/02, Issue # 801, page 5.
