SI
SI
discoversearch

We've detected that you're using an ad content blocking browser plug-in or feature. Ads provide a critical source of revenue to the continued operation of Silicon Investor.  We ask that you disable ad blocking while on Silicon Investor in the best interests of our community.  If you are not using an ad blocker but are still receiving this message, make sure your browser's tracking protection is set to the 'standard' level.
Technology Stocks : Rambus (RMBS) - Eagle or Penguin -- Ignore unavailable to you. Want to Upgrade?


To: unclewest who wrote (18193)4/3/1999 9:32:00 AM
From: MileHigh  Respond to of 93625
 

April 05, 1999, Issue: 1154
Section: News
--------------------------------------------------------------------------------
Intel outside?
Jack Robertson

Ironically, Intel Corp. in a few cases is falling victim to its co-founder's maxim, Moore's Law.

The microprocessor giant just didn't count on how quickly the market would be blitzed with the alternative types of memory it had rejected. For now, this has created the novel situation of Intel steadfastly holding on to older technology that others may have surpassed.

The surprisingly fast ramp of the PC-133 SDRAM that Intel spurned is starting to give the chip maker a little heartburn. Years ago, when Intel saw that next-generation processors would need super-speedy memory, it selected Direct Rambus DRAM-the most advanced wideband memory architecture at the time. Few, including Intel, could foresee that 133-MHz SDRAM would develop so rapidly. PC-133 memory and its successor, double-data-rate SDRAM, now pose a serious challenge to Intel's game plan.

Intel now finds itself campaigning against PC-133 instead of considering new technology. An Intel spokesman expressed skepticism that the loose organization of PC-133 backers-memory companies, chipset makers, motherboard builders, connector manufacturers, PC OEMs-could come up with the complex unified specifications to make PC-133 operate stem-to-stern inside the box. After all, Intel knows how daunting it was to get PC-100 memory to work.

But an IBM Microelectronics spokesman said the intricacies of PC-133 were worked out more expeditiously than those of the earlier PC-100 effort largely because of the little-known fact that IBM had worked jointly with Intel on the PC-100 specifications. "We could build on this base of experience and bring PC-133 to market at a pace that no one ever thought possible," the spokesman said. Certainly Intel didn't.

Similarly, EDO DRAM, which was state-of-the-art when Intel adopted it as the sole memory for its Pentium II/III Xeon server chips, is now rapidly being phased out of mainstream production by suppliers. This is a twist: Intel, which prides itself on staying at the leading edge of technology, is now dependent on a trailing-edge memory technology for its most advanced processors.

Due to delays in getting Direct RDRAM and the allied Camino and Carmel chipsets to market, Intel is now in the embarrassing position of being outgunned this summer by faster chipsets from independent vendors Via Technologies, Reliance Computer, and perhaps others.

All this could change when Intel and Direct RDRAM suppliers break forth on the market later this year. The 800-MHz Direct Rambus memory linked to a 500- to 600-MHz Pentium III processor could be a true Dynamic Duo.

Meanwhile, however, Intel's persistent refusal to acknowledge the blazing progress of alternative types of memory, for whatever reason, has created a take-no-prisoners technology battle-and it probably need not have happened.

Copyright ® 1999 CMP Media Inc.





To: unclewest who wrote (18193)4/3/1999 9:39:00 AM
From: MileHigh  Read Replies (1) | Respond to of 93625
 

April 05, 1999, Issue: 1055
Section: News
--------------------------------------------------------------------------------
DRAM slide drives rush to high ground
Anthony Cataldo and Yoshiko Hara

Tokyo - Groping for fresh ways to stem massive losses in DRAM sales during the past year, several of Japan's chip makers are gambling that they can shield themselves from further hits by focusing on higher-density, leading-edge DRAMs. Whether they can do so in an environment of consolidation for a commodity product remains a question.

The rethinking of DRAM strategy came to light amid a week of bloodletting and reorganization in Japan as the memory market, which had been flirting with stability for a short time, resumed its two-year slide.

As Japan's fiscal year drew to a close on March 31, the country's DRAM vendors reaffirmed their intentions to keep capital spending flat even as their market share has plummeted.

South Korean companies, by contrast, have gained considerable market share and are upping their capital spending 15 percent above last year's total. The Korean vendors cut capital budgets 65 percent in 1997 and 30 percent in 1998, said Ilsuk Han, an analyst with ING Barings Securities Korea.

Rather than add more capacity, Japan's chip vendors are looking to upgrade existing fabs with leading-edge equipment while turning their energy toward newer product lines, such as 128-Mbit DRAMs, double-data-rate (DDR) models, Direct Rambus and 133-MHz synchronous DRAMs.

Toshiba Corp. announced last week it would cut back 64-Mbit DRAM production 90 percent by year's end, shifting the majority of its DRAM output to 128-Mbit SDRAMs and high-performance Direct Rambus and DDR parts.

The company's sales revenue plunged more than 50 percent between 1997 and 1998, moving from $1.6 billion to $730 million, according to Semico Research Corp.

Stephen Marlow, vice president of memory marketing for Toshiba America Electronic Components (Irvine, Calif.), said the company is leveraging its 0.2-micron process-the thinnest production-level geometry in use-to push out 128-Mbit and denser parts in hopes of improving margins.

Marlow blamed Korean vendors for resuming mass production in late December, leading to a return of the DRAM price wars in late January.

At the end of last year, "people said, 'We can make money at $10 apiece; let's pump some more [production] through.' Now we're looking at a feeding frenzy," Marlow said. "If we don't act maturely as an industry, we're going to see a retrenchment and a consolidation, and it will impact our future technologies."

Fujitsu Ltd. has likewise said it will tilt toward high-density, leading-edge products. Ryusuke Hoshikawa, chairman of Fujitsu Microelectronics Inc., said in an interview that the company's Fast-Cycle RAMs and other exotic memory blends will get more of its attention in the future. The company is shifting memory production at a fab at Iwate, Japan, to logic, and last week Hoshikawa unveiled a Web-based design environment to make it easier and faster for customers to get system-on-chip logic designs from Fujitsu.

Semiconductor operations at Toshiba and Fujitsu, both of which have lost considerable market share, are under the gun as electronics vendors in Japan gird for further pressure to return to profitability at the close of a disastrous fiscal year.

But is abandoning the so-called commodity DRAM market a realistic plan? "That's the million-dollar question," said Richard Kaye, an analyst with Merrill Lynch Japan. "Right now, Toshiba is just one of many companies trying to de-commoditize DRAM products."

The turmoil is helping to make DRAMs an increasingly fragmented technology, as new speeds and memory types come on stream. Meanwhile, some vendors see spot opportunities in trailing-edge technology.

Mitsubishi Electric, for one, has agreed to extend production of 4- and 16-Mbit devices for some customers, even though it had earlier planned to phase out such parts, said Cecil Conkle, assistant vice president of DRAM marketing for Mitsubishi Electronics America Inc.

"Our plan was to exit those products, but as we go through the steps to do that there are some critical customer requirements that have us extending [them] longer than we had planned," he said.

And why not? Conkle noted that contract prices for 64-Mbit DRAMs are falling below $10 this year amid weakness in PC demand in the United States. At the same time, contract prices for a 4- or 16-Mbit device are edging up to $5 apiece (against less than $4 in the spot market).

As it wrestles the DRAM demon, Mitsubishi's top management is downplaying the role of DRAMs in the company's future. "In three years, we plan to halve the percentage of commodity DRAM in total sales," Ichiro Taniguchi, president of Mitsubishi Electric Corp. said when the company announced its midterm corporate strategy last week. Taniguchi predicted that Mitsubishi's semiconductor business will return to profitability in fiscal year 2000.

Mitsubishi positions two sectors-microcontrollers and embedded-DRAM products, a category it calls eRAM-as its competence fields.

Even as the Japanese titans sketch plans to minimize their reliance on DRAM sales, other Asian manufacturers are all too willing to pick up the slack. Taiwanese DRAM vendors are pushing processes to wring more profits from their lines.

Toshiba said last week that it will transfer its 0.15- and 0.17-micron CMOS process to Winbond Electronics Corp. in Taiwan. And Nan Ya Technology Corp. there said it is getting process technology from IBM that will allow it to skip the 0.25-micron manufacturing step and begin with 0.2-micron technology to produce 64-Mbit DRAMs.

Profit from process

The potential for profit at the lower geometries is impressive. About 350 of the 64-Mbit dice can potentially come from one 8-inch wafer using quarter-micron process technology. At 0.2 micron the potential yield goes to 550 dice; at 0.17 micron it jumps to 800.

Nevertheless, the vagaries of the market and the suffering it has brought suppliers is leading Japanese DRAM vendors to restructure their business units. Several companies are looking ahead to the time they will no longer be able to count on the corporation to prop up unprofitable businesses.

In the past, for example, Japanese chip companies would often cross-subsidize semiconductor capital spending by diverting revenue from telecom equipment sales to national carrier NTT. Companies would use the money to help fund next-generation DRAM development, while keeping profit margins thin. From 1987 to 1997, Japanese vendors' average profit margin was 3 percent, said Kaye of Merrill Lynch.

That practice is coming under fire, however, as NTT slashes its own capital budget this year. One ray of hope is cultural-the Japanese reluctance to fire workers appears to be softening. U.S. observers believe a judicious program of downsizing would go a long way toward cutting overhead and fostering more nimble operations.

"In the next year or two, you'll see significant cuts. There's been a tremendous change in the operating thought process," said analyst Handel Jones of IBS Inc. (Los Gatos, Calif.). -Additional reporting by Brian Fuller.

Copyright ® 1999 CMP Media Inc.





To: unclewest who wrote (18193)4/3/1999 10:20:00 AM
From: REH  Read Replies (1) | Respond to of 93625
 
Future of Alpha chip revealed

Terry Shannon, editor of newsletter Shannon knows Compaq is delivering a presentation today at a user group meeting which reveals the future of the Alpha roadmap.

Shannon is an analyst and journalist who has tracked Digital and latterly Compaq for years.

He has kindly sent us a copy of the presentation he will make to DECUS today, and interesting reading it makes.

He says that the Alpha 21264 EV6 delivers blistering speed and the EV67 will be available shortly. He thinks that it will clock at 700MHz for TurboLaser and 767MHz ins for WildFire, attaining an estimated 36Specint95 at 667MHz.

By the end of the year, the EV68 will arrive at speeds of around 833MHz, and possibly 1GHz by year end.

According to Shannon, the 21364 EV7 Alpha will use the EV6 core plus on-die level two cache and support logic. It will provide what he describes as "glueless" symmetric multiprocessing (SMP), RAMBUS support, lock stepping for NSK.

The EV8 21464 Alpha "Arana" technology is expected to arrive towards the end of 2001, in the same time frame as Intel's McKinley architecture is due. It will offer symmetric multithreading and 200SPECint95 at 1200MHz and even 1400MHz.

The EV9 and EV10, thinks Shannon, are in advanced stages of development.

Shannon says we will see one to four CPU Regatta class ES40s in Q2 of this year, along with the WebBrick DS10, which will have one to two CPUs.

Again, in the same timeframe, we will see the VMSstation VS10.

Compaq will release a 1-2 CPU Alpha ProLiant called the Catamaran in the middle of this year along with a one to four CPU ProLiant dubbed Schooner.

We will see WildFire 16-CPU boxes by year end, as earlier revealed here.

Future NT ProLiants will include the eight CPU Python, Diamondback, Cobra, and Copperhead boxes in the first half of next year. ®



To: unclewest who wrote (18193)4/3/1999 1:11:00 PM
From: Dave B  Respond to of 93625
 
I bookmarked the article -- I'll read it later. Off to a birthday party for twin 2-year old boys right now. Yahoo.

Dave