Programmable Notes

Going for the Jugular It's a new wrinkle in chip marketing, but the question is, will it work? By Larry Waller

Even the most grizzled semiconductor industry veterans, who have seen it all, say this tack is a new one on them.

What grabbed their attention during the holiday season is the spectacle of one company publicly attacking a rival's new product line at some length and with a gusto many found unseemly. AT&T Microelectronics (Allentown, PA) is the aggressor company, going after San Jose, California-based Xilinx's new XC5000 family of FPGAs no-holds-barred.

AT&T shocked the industry by issuing a press release in late November, just one day after Xilinx launched an important addition to its FPGA lines amid much fanfare. In what amounted to a full broadside, AT&T challenged the validity of a number of claims made by the FPGA leader about architectural improvements offered by the XC5000 products.

The critical tenor of the AT&T release, and particularly the subsequent statements of FPGA product director Chris DeMonico, left no doubt that the company had its competitive instincts cranked up far higher than the usual industry practice. In a time-honored tradition, new product announcements are often regarded as the occasion for give and take between programmable logic rivals who make competitive selling points; but it is essentially good-natured. AT&T's response, on the other hand, went considerably beyond the mostly gentlemanly marketing banter that in the past has characterized these exchanges.

One factor that elevates the AT&T- Xilinx affair above the norm is the significance of the products involved. Xilinx has an ambitious target with its 0.6µm XC5000 FPGAs; nothing less than becoming the standard programmable logic element to fill the 10,000 to 50,000 gate density range. Pickings in this market are expected to be rich for the programmable logic supplier who gains the edge, since gate array manufacturers are candid about pulling out of a no-profit business to concentrate on the very-high density devices. As the FPGA leader, Xilinx fully expects to extend its dominance by powering its new products into the void as gate array replacements.

This strategic move obviously hits a raw nerve at AT&T, whose ORCA FPGA already has chalked up the best initial success for design-ins, on the strength of products offering 12,000 and 26,000 gate densities. AT&T is in the position of having its flagship ORCA line threatened by the efforts of a company that now holds 40 percent of the FPGA business. Because its total sales are far smaller than Xilinx's , it no doubt feels that mounting a very combative response is most effective.

For openers, AT&T's DeMonico charges that Xilinx's XC5000 family brings the market "nothing substantially new.... [it] embodies architectural techniques previously introduced in competing products." And these competing products are ORCA devices that have been on the market for several years.

These me-too developments are the VersaBlock Logic Module, which combines logic functions with local routing to gain higher efficiencies, and the VersaRing, which decouples I/O blocks from core logic for better pin assignment flexibility. "Both of these techniques were first seen with the ORCA 1C Series over two years ago," DeMonico says. Moreover, the Xilinx parts do not have on-chip user RAM as does ORCA, which he calls essential.

But DeMonico saves his most telling criticism for the all-important speed specifications. He claims that FPGAs produced with the 0.6µm Xilinx process do not keep up with AT&T's 0.5µm triple-layer-metal devices, which are the only ones of this type in volume production. Having sufficient programmable device performance to replace speedy gate arrays has been somewhat of a problem in the past, but AT&T's newest ORCA, the 2C Series, has specs that include a 3.8ns lookup table (the basic computational function) and 68MHz fast carry logic. This and other features make it 25 percent faster than the new Xilinx line, according to the AT&T official.

Besides the performance features, AT&T has gotten ahead of rivals Xilinx and Altera (San Jose, CA) in offering the higher density devices, in the view of DeMonico. The firm broke the 10,000-gate barrier two years ago, is now sampling a 26,000 gate FPGA, and expects to debut a 40,000-gate part this quarter. In the meantime, Xilinx will have its 10,000-gate XC5215 this quarter, and the 15,000-gate XC5215 later in the year. Altera does not make FPGAs but its high-density CPLDs compete for many of the same sockets; its 16,000-gate FLEX 8000 CPLD has been on the market since last year, and is reportedly doing well. "We're going after bigger game than 10,000 gates," says DeMonico. "It's a heck of a lot less crowded above 10,000 and we have the only viable product."

Despite all the AT&T flak, Chuck Fox, Xilinx's vice president of product marketing, is not replying in kind. "No tit for tat," he says. Instead, he downplays any lasting significance. "It reflects [AT&T's] frustration" in taking on rivals who are well-established in the PLD business, he says. Moreover, AT&T has "a short-sighted strategy on ORCA, which it claims can do everything. Our strategy is completely different."

Xilinx is committed to multiple architectures, and the XC5000 is the latest, with two more on the way during the year. The real issue is how the customer will accept the products, and the buyer is "not swayed by competitive claims" such as the broadside leveled by AT&T. On this score, Xilinx believes the new family will be a winner, and Fox is "very confident that the architecture is efficient."

As to how to view the AT&T attacks, Fox advises "it might be a mistake to read too much into this. But issuing a press release to refute another company's products point by point does show a lack of professionalism."

Altera is not exactly a disinterested observer in the AT&T-Xilinx skirmish, since DeMonico also consigns it to an also-ran in upper-level CPLD density. But Altera is not the principal target, either. Still, Bob Beachler, Altera's manager of strategic marketing, thinks AT&T has taken the low road; "they're out there bashing Xilinx ."

While not approving of the AT&T gambit, Beachler's firm does compete strongly with Xilinx , so he can't resist throwing a barb: "the XC5000 is open for attack. It brings nothing to the game." The pricing issue is especially misleading, he says, because Xilinx is promising a tab of $39 for its 10,000-gate XC5210 when volume production is reached this quarter. Altera's 12,000-gate 8188A pricing is already there, and will drop further soon.

As for the effectiveness of AT&T's claims, "customers will prove them or not," Beachler predicts. This opinion is the consensus of most observers on the outcome of AT&T's attack marketing. But if the strategy works well, they say, the industry will see more of it.

To most observers, the high-decibel marketing volleys among programmable leaders further confirms the view of market analyst Rhondalee Rohleder of Pace Technologies (Scottsdale, AZ), who says that at higher PLD densities, "competition is very intense." AT&T's DeMonico is even blunter. "It's a battle royale for sockets," he says, "and it's a brand new ball game."

Larry Waller is a contributing editor of Integrated System Design.

To voice an opinion on this or any Integrated System Design article, please e-mail your message to: michael@asic.com.

integrated system design  March 1995