CNET también está disponible en español.

Ir a español

Don't show this again

Christmas Gift Guide
Tech Industry

Will K7 be the breakout chip for AMD?

The upcoming K7 could help AMD conquer new markets--or could be the latest in a line of processors that AMD has trouble producing in quantity.

The upcoming K7 processor could help Advanced Micro Devices break out of the low-end PC segment, loosen Intel's virtual monopoly in corporate computing, and inaugurate an era of profitability and technical respect for the chipmaker.

Then again, it could also be the latest in a series of AMD processors that are tough to find, difficult to manufacture in quantity, and ultimately relegated to discount pricing.

The fate of the K7, which debuts in June, will become one of the major issues for the processor and PC industries for 1999. Nearly all analysts agree that, on paper, AMD has come up with a powerful, evolutionary processor that in many respects will outperform the Intel Pentium III, the company's standard bearer in the business market, at a lower cost.

AMD's chip could speculatively even become a competitor to Intel's high-end Xeon processors for servers by eliminating $20,000 or more from the manufacturer's cost, according to some.

The chip will come out at 500 MHz or faster, according to AMD. But the company today demonstrated a version of the chip running at 1 GHz (1,000 megahertz) in a supercooled computer from Kryotech at AMD's stockholder's meeting in New York.

"We believe that we will be able to have the highest performance Windows-compatible processor for the next six to 12 months," boasted chief executive W.J. "Jerry" Sanders earlier this month.

On the other hand, nearly all observers note that AMD has a long and difficult history when it comes to product releases. AMD has lately consistently missed its manufacturing and financial goals. To top it off, the K7 is one of the most ambitious projects the company has ever tackled.

AMD officials are confidently optimistic, but acknowledge the scope of the undertaking.

"The K7 is a totally different animal," compared to existing X86 processors, stated Lance Smith, director of technical marketing at AMD. "The K7 is designed to do high frequency, period."

Nonetheless, he added, "The problem is that [Intel has] eight fabs [chip fabrication facilities] in parallel. One croaks off, big deal. We have one fab. If we hiccup, everyone knows."

Kudos for chip design
Although the ultimate commercial destiny of the K7 remains unclear, analysts have generally applauded the chip's design. Roughly, the design improvements can be categorized as "more and better." The K7 can handle more instructions per clock cycle than the K6-2 and can process these instructions more efficiently than the Pentium III, according to AMD and others.

AMD has also eased back on the speed at which the transistors switch inside the processor, said Nathan Brookwood, principal at Insight 64. With the K6, AMD pushed transistor speed, which ultimately limited the clock speed of the chip family.

"Their design doesn't push so hard on transistor speed, so they can incrementally improve it," he said. Overall, "if it all comes out right, with a K7 at the same clock speed as a K6-2, they should have substantially better performance."

In addition, the K7 will use the EV6 system bus typically associated with the Alpha processor Compaq. The EV6 connects the chipset to the processor at 200 MHz, twice as fast as the bus on the Pentium III.

"This will bring an ability to get a lot more data into the processor," said Brookwood. "Intel is constrained on some of the benchmarks because of the bus."

Commercial intrigue
That sort of performance opens up the commercial market for AMD, which could create difficulties for Intel, according to many.

The K7 will likely come out in the $400 to $500 range in wholesale quantities, according to estimates. Although expensive by consumer-chip standards, it is lower than the cost of the fastest Pentium III processors, which sell in the $700 range. Historically, AMD has maintained lower prices than Intel. Intel, however, has battled back by cutting prices on low-end chips, accelerating development to the point where AMD began to lose money, and recovering profits with higher-end chips. Assuming the K7 succeeds, the question is how far Intel will be able to cut its prices and margins on the Pentium III.

"This puts Intel in an interesting position," said Dean McCarron, principal at Mercury Research. Intel may "have to focus on something other than price," he added.

AMD could become an even bigger threat if they could manage to break into the server and workstation space, according to A.A. LaFountain, semiconductor analyst at Needham & Co. Currently, Intel charges around $3,600 for the 500-MHz Pentium III Xeon processor with 2MB of secondary cache. Most of that, however, is profit. The 1MB version costs just under $2,000 and the difference in cost between the two chips is probably less than $100, he said.

AMD could replicate the performance of the top Xeon processor at around $500, or close to $3,000-less per chip. In an eight-processor server, that comes to a reduction in manufacturer's cost of $24,000.

And, while an eight-processor AMD system may sound farfetched, a company called Poseidon Technology is coming out with a K7-compatible chipset that will make that possible. "We've discussed the concept with most of the major computer makers. We've gained a level of interest," said Rick Shriner, chief executive of Poseidon. Poseidon will demonstrate the chipset at Comdex later this year.

On a more mundane level, the K7 could also prevent erosion on the K6-2 and K6-III line of consumer processors. With the new chip, AMD can potentially begin to segment its product line, AMD vice president of marketing Dana Krelle has said.

Will history repeat?
History, of course, does not favor AMD. Over various successive quarters in the past two years, the company has admitted to problems in meeting manufacturing goals with the K6, the K6-2, and the K6-III. Financially, success with the K7 is imperative. AMD has set a goal of selling its processors for an average price of $100, and Sanders recently said that the company "needs the K7" to hit that mark.

"The K7 is a pretty fantastic design," said Kelly Spang, an analyst at Technology Business Research. "But for whatever reason, its product lines have been inconsistent."

A major hurdle the company has faced in the past has been a lack of factories. Currently, the company has one. If the manufacturing process is not perfect, problems emerge. A second will open in Dresden, Germany, later this year, but overall factory capacity will still be dwarfed by Intel. Oddly, AMD could get more factory capacity through its deal with IBM, noted Spang, but AMD has yet to exercise the right.

In addition, chipset and motherboard makers have to supply parts to make this happen. "The K7 needs to have chipsets in place," said McCarron. Companies have agreed to build these parts, but, again, how it plays out in the marketplace remains to be seen.

Finally, the company will have to overcome inertia. "I wouldn't expect anyone to run off and do an 8-way K7...[but] if they make it easy for someone to put together a dual-processor workstation, people will be interested," said Brookwood. Nonetheless, "The business PC market is somewhere where they have some reputation-building to do."