The Microventures conference, which features hordes of startup CEOs pitching their companies to bored-looking venture capitalists, is a great place to check out the leading edge of chip technology. Some of these startups are closer to the edge than others.
TCP offload has become a hot topic now that some Gigabit Ethernet connections are approaching their full capacity and 10G Ethernet is on the horizon. At these speeds, the TCP stack alone can bog down even a high-frequency Pentium.
NetEffect is one of several startups developing chips to offload TCP at up to 10 Gbits/second using a new Windows standard called iWARP. Starting in the first quarter of 2005, NetEffect plans to sell its chip as part of a full channel-adapter board.
Major Ethernet vendors Broadcom, Intel and Marvell have yet to deploy custom silicon for TCP offload. Broadcom has been active in iWARP, but Intel, not surprisingly, continues to recommend faster Pentium processors for this task. The delay leaves room for one or more startups to succeed.
Tarari is a leader in the emerging market for XML acceleration. The startup has developed an FPGA-based board that can be configured for XML parsing and grammar analysis at speeds of up to 2 Gbits/s.
The flexible FPGA approach helped Tarari ship its first hardware nearly two years ago, and it has been fine-tuning its algorithms since. The company plans to convert its FPGA into an ASIC in the first quarter of 2005, improving performance and reducing cost.
The startup showed a Zapthink study estimating XML makes up 10 percent of network traffic today-and that could hit 40 percent by '08.
Broadlight is a leading vendor of passive optical networking (PON) chips, another tiny market that could grow fast. PON provides a low-cost way to drive a high-bandwidth fiber directly to the home or business. Startup Passave also is focused on PON. SBC alone plans to deploy PON to 1 million homes in 2005.
On the far edge are nanoelectronics startups such as ZettaCore. Instead of traditional trench capacitors, the company uses custom-designed molecules to hold a charge. It has fabricated a test chip that demonstrates its molecular memory.
Will ZettaCore change the world? That remains unclear, but any one of these small companies could become the next big thing.
Linley Gwennap is founder and principal analyst of The Linley Group and co-author of "A Guide to Storage Networking Silicon" (www.linleygroup.com/npu).