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: Mihaela who wrote (66872)3/1/2001 10:26:46 PM
From: Don Green  Read Replies (1) | Respond to of 93625
 
Intel to integrate USB, wireless into P4 chipsets
By: Tony Smith
Posted: 01/03/2001 at 18:58 GMT

Intel Developer Forum, San Jose Intel will add native USB 2.0 support to its PC system chipsets early next year, the company's desktop products group chief, Louis Burns, promised today.

And wireless 802.11 support will follow around a year later, he pledged.

Intel's roadmap highlights the release during Q1 2002 of a version of its Brookdale chipset, which will debut early this summer, designed to support DDR SDRAM. Certainly, there's nothing to suggest Brookdale will offer native USB 2.0 support, so we suspect that it will come in with the successor to today's RDRAM-based P4 chipset, the 850.

But whatever chipsets ends up offering USB 2.0 support - and we reckon if one does, they all will sooner or later - it will certainly help drive the adoption of the upgraded, 480Mbps peripheral interconnect technology.

Essentially, it's part of Intel's strategy to drive the Pentium 4 platform as the core for rich media enabled systems - its digital convergence platform, if you will. The P4 can't flourish as a multimedia host unless Intel makes it as easy as possible for users to get high quality video and audio into their systems.

Fans of IEEE 1394 (aka FireWire and iLink) will wonder at this point whether Intel will add native support for that technology to its chipsets. Burns' answer will disappoint them: 1394 will not be supported natively, he said. Instead, the company is working to get wireless 1394 to hook into a PC system via wireless connections.

Burns was fairly positive about 1394, given some of the brickbats his colleagues have previously whacked it with in the past, but the bottom line is that Intel's stance on the technology remains unchanged: 1394 is for consumer electronics applications, USB 2.0 for PC-based roles.

Native wireless support will bridge the gap, allowing Intel-based PCs to live in a 1394 world without having to participate in it directly. In other words, Intel bringss all the benefit of 1394's wide CE industry support to the PC without actually having to go back on its words and admit that 1394 has a part to play in PC design. That's as good a way of getting cake and eating it to as we've seen in a long while.

Burns said Intel was working hard to promote 1394 over wireless and to drive the standard, though it has no plans to build silicon. The 802.11 side of the equation is interesting too, since Burns' IDF keynote demo was up in the 5GHz band, suggesting Intel wants 54Mbps 802.11a to replace the slower 911Mbps), lower frequency 802.11b that's powering most RF networks at the moment.

The chip giant is also driving the development of 802.11e quality of service extensions to both 802.11a and 802.11b. Improved QoS is essential to the delivery of high quality rich media over wireless, and that's particularly important if your strategy revolves around getting that data to your processor technology. ®



To: Mihaela who wrote (66872)3/2/2001 2:40:02 PM
From: Don Green  Read Replies (1) | Respond to of 93625
 
Intel Invests In RDRAM Capacity
By Steven Fyffe, Electronic News
Mar 02, 2001 --- SAN JOSE, Calif.— Intel Corp. wants more Direct Rambus (DRAM), and it is willing to pay extra for it. Samsung Electronics Co. Ltd. has accepted an undisclosed sum from Intel in return for a promise to boost RDRAM output, the Korea-based company revealed at the Intel Developer Forum this week.

Samsung will start cranking out 10 million RDRAM units a month immediately, and has pledged to increase that figure to 20 million a per month in the second half of this year. Samsung plans to use Intel’s money to buy high-speed RDRAM testers.

“The wafer capacity for Rambus was always in place,” said Dieter Mackowiak, senior vice president of sales and marketing at Samsung Semiconductor Inc. “The bottleneck was the back-end testing, because the Rambus test is more intensive. With this agreement, Intel is going to invest an unspecified amount of money into Samsung, and Samsung is going to invest in test equipment.”

Industry insiders estimated that Intel (nasdaq: INTC) spent $100 million to $200 million on the Samsung deal. Geoff Tate, chief executive officer of Rambus Inc., previously told analysts that it takes roughly $10 million worth of test equipment for every 1 million units of RDRAM tested each month, one source said. RDRAM testers sell for $5 million to $6 million apiece, said Ron Leckie, industry veteran and chief executive officer of INFRASTRUCTURE market research.

Intel has approached other DRAM makers with similar deals. Elpida Memory Inc. has rejected the offer so far. The company doesn’t want to be tied down to the strict pricing and volume requirements that Intel is demanding.

“If it was free money, of course we would accept that,” said Hidemori Inukai, vice president and general manager of the technical marketing division at Elpida, Tokyo, Japan.

“In the past we have not accepted, and probably this time we will not accept,” he said. ”We always welcome free money, but Intel is looking for something in return for the investment. At least some kind of insurance they have to have, about price and certain conditions they are asking about.”

Santa Clara, Calif.-based Intel is determined to make sure there is enough RDRAM to support the aggressive rollout plan for its Pentium 4 processor. Intel is even buying some RDRAM itself and offering cut-price package deals to white-box makers.

“There’s more than sufficient RDRAM capacity locked up to meet the early ramp of the P4,” said Paul Otellini, executive vice president and general manager of the Intel architecture group. “We’ve got agreements from a number of the major manufacturers and Intel is purchasing some for bundling with motherboards that go into the white-box channel.”

At IDF, Intel rallied the faithful DRAM makers to present their memory roadmaps, including Samsung, Elpida and Toshiba Corp. Silver sponsors of IDF — Hyundai Electronics America and Micron Technology Inc., — which are both suing Rambus (nasdaq: RMBS) for alleged antitrust violations, were notably absent from proceedings.

Despite its refusal to take Intel’s offer, Elpida said RDRAM would make up 30 percent of its total output by the end of the year. And Toshiba announced plans to triple RDRAM production by the third quarter of this year.

The increased volumes and a cheaper four-bank architecture that should be coming out early next year will help bring RDRAM prices down, said Intel Fellow Pete MacWilliams.

“The cost structure of RDRAM will be very competitive long term. The 850 chipset isn’t the last Rambus chipset we have planned,” MacWilliams said. “We have another one planned for 2002, and it will include support for the 4i structure.

“We also have an SDRAM solution for the P4 in the second half of this year, and we will have a DDR platform in the first half of next year. We view this as another price-performance option.”

MacWilliams glossed over Intel’s love-hate relationship with Rambus that has played out in the press over the last year, and assured the crowd that Intel was still committed to RDRAM.

“There is no fundamental change to Intel’s position on memory. We have had to make some technology decisions that keep making people second-guess our intent. I’m here to reinforce; this is not the case.”

He also downplayed the potential threat of double data rate (DDR) memory to the future of RDRAM.

“If the RDRAM pricing plays out, the role of this DDR could be extremely limited,” MacWilliams said. “While there will be a robust DDR platform next year, there is a robust RDRAM platform now … and that’s what we are expecting to drive the P4 ramp,” he added. “How that plays out in 2002 is anyone’s guess. All the people we have enabled have incentives to hit the market.”

Samsung claims an RDRAM module could cost only 5 percent to 10 percent more than a DDR module at the end of the year.

“Eventually, when DDR and RDRAM become high volume, we believe that RDRAM’s theoretical cost can approach about 5 percent of DDR,” said Jon Kang, senior vice president of Samsung’s memory division.

Samsung’s prediction was not likely to come true, Elpida’s Inukai said. “There is still no (4i) chipset to support it, so it is not a realistic number,” Inukai said.

The current 2 x 16d architecture probably costs 20 percent to 25 percent more than SDRAM and 10 percent to 15 percent more than DDR, he said.



To: Mihaela who wrote (66872)3/2/2001 2:40:54 PM
From: Don Green  Read Replies (2) | Respond to of 93625
 
DRAM Feast Could Turn To Famine
By Steven Fyffe, Electronic News
Mar 02, 2001 --- The current glut of cheap DRAM could quickly swing to patchy shortages and higher prices as manufacturers juggle their product mixes and the battle between direct Rambus DRAM (RDRAM) and double data rate (DDR) memory heats up.

“In the April-May timeframe we will first start to see the signs of switching to next-generation memory,” said one industry veteran who now works for Infineon Technologies AG, Munich, Germany. “As companies adjust their product mix, there will be shortages all over the place.”

Shortages would be a blessing for DRAM makers plagued by plummeting average selling prices since the end of last year. The slowing U.S. economy could even be a good omen for DRAM companies that have traditionally bucked the prevailing macroeconomic trend, the source said.

“The memory industry has got a long history of having good times while the national economy is in bad times,” the source said. “I have no idea why that is, but it has been historically correct for about 30 years now.”

Many DRAM makers are now losing money on certain types of SDRAM. The rush to offload unprofitable product lines could cause some specific shortages, said Dieter Mackowiak, senior vice president of memory sales and marketing at Samsung Semiconductor Inc., a subsidiary of Samsung Electronics Co. Ltd.

“There will be a structured shortage,” Mackowiak said. “Everybody is trying to reduce 128Mbit SDRAM as much as possible because it’s not a profitable product right now, like EDO or RDRAM.”

Demand for RDRAM could also outstrip supply this year, even with Samsung, Elpida and Toshiba pledging to boost production, Mackowiak said.

Others said shortages could be more widespread and difficult to predict.

“That’s certainly a potential,” said Gary Swanson, senior vice president and general manager of the DRAM division at Hyundai Electronics America, the U.S. subsidiary of Icon, South Korea-based Hyundai Electronics Industries Co. Ltd. “It could effect some of the higher-density products if the forecast isn’t there. Even traditional DRAM, such as PC-133 SDRAM, could be impacted.”

It wouldn’t take much demand pressure to swing the market back to undersupply, Swanson said.

“A lot of inventory is going to be burned off by the end of March,” he said. “The portable industry is really taking off, and our large PC customers, their inventories look to be at reasonable levels. And Hyundai’s inventory is not very large. We don’t have two-months’ worth of inventory. If the market picks up at all, you could start seeing mix issues start to surface.”

Others were skeptical of the shortage theories, labeling them wishful thinking.

“I wouldn’t say we’re seeing any kind of shortages,” said Jeff Mailloux, director of DRAM marketing at Micron technology Inc. in Boise. “We’re just trying to get our mix to match up with what our customers are expecting. I don’t really foresee any big issues with the DDR ramp. The DDR ramp is going pretty smoothly and there’s a lot of flexibility there. In terms of total DDR supply and demand, there’s pretty good supply vs. demand. It’s a matter of getting good forecasts from the customers. But there may be some short-term mix issues.

“There are some customers that are starting to talk about longer-term plans and getting the right mix of DRAMs. The fact that people are starting to think about their supplier for the fall is a good sign.”



To: Mihaela who wrote (66872)3/2/2001 8:22:35 PM
From: Don Green  Respond to of 93625
 
Intel has spotty record luring DRAM makers to build Rambus chips
By Jack Robertson, EBN
Mar 2, 2001 (4:43 PM)
URL: ebnews.com

Desperate to get enough Direct Rambus DRAMs into the market to support its new Pentium 4 processor, Intel Corp. this week announced it has put money into Samsung Electronics Co. Ltd. for the second time to boost RDRAM output and ensure a steady supply of parts. Samsung appears to be the only taker of Intel's Rambus-related investment offers. Both Hyundai Electronics Industries Co. Ltd. and Elpida Memory Inc. said they have turned down Intel investment offers.

Based on past experience, however, Intel's desire to swap cash for a promise of chips may not pay off. To date, the company has been unsuccessful in its attempts to motivate other chip companies to start making RDRAMs or boost production sharply.

A total of $750 million that Intel has invested in Micron Technology and Infineon Technologies, which the chip maker said was intended to spur RDRAM efforts at these companies, has failed to bring them on line.

Samsung makes 70% to 80% of the industry's RDRAM supply, according to its own estimates. A year ago, Intel invested $100 million in Samsung, which it said was to spark RDRAM production.

Samsung will use this round of Intel money to install additional advanced testers so it can ensure higher RDRAM output. The Korean company had been producing more Direct Rambus die than could be tested, creating a back-end logjam, according to Dieter Mackowiak, senior vice president of sales and marketing at Samsung Semiconductor Inc. in San Jose.

"We can install the new testers quickly with the Intel investment, especially since we can get equipment from orders canceled by other chip makers," Mackowiak said.

But Intel needs to secure additional sources among the other large Rambus producers if it expects to meet its current Pentium 4 needs, according to analysts.

The first-generation Pentium 4 Willamette chip is locked into using Direct Rambus memory only, a legacy of Intel's prior policy to move to RDRAM exclusively for its advanced desktop processors. "That leaves Intel in dire straits to get memory suppliers to build enough RDRAMs to support the aggressive Pentium 4 ramp-up," said Nathan Brookwood, an analyst at InSight 64, Saratoga, Calif.

"Without enough Rambus DRAMs in the market, the Pentium 4 ramp is constrained," Brookwood said.

Besides Samsung, Intel appears to have struck out with other chip makers on this round of funding.

Sang Park, president of Hyundai Electronics' Semiconductor/LCD group, told EBN last month that Intel made the RDRAM money offer to his company first, but the two sides could never come to an agreement.

Similarly, Hidemori Inukai, Elpida Memory's vice president and general manager of the technical marketing division, said the NEC-Hitachi joint DRAM venture failed to agree on an Intel money proposal.

Even without the Intel funding carrot, Elpida is boosting its Direct RDRAM output from roughly 3 million 128 Mbit equivalent chips in the first quarter of 2001 to 18 million to 20 million in the fourth quarter, Inukai said.

Intel failed to bring on Micron Technology as a major RDRAM supplier as it had intended by investing $500 million in the DRAM maker in 1998. Micron president Steve Appleton said the company was free to use the Intel investment as it wished. Micron has developed and validated its own RDRAM chip, but still claims it is waiting on market demand before going into production.

The Micron investment, however, turned out well for Intel in another way: the MPU company reportedly made nearly $1 billion from reselling the Micron shares it had acquired.

Intel's $250 million investment in Infineon last year has also failed to bring the German chip maker into the RDRAM production ranks, despite Intel's spin at the time that this was an intended purpose.

Infineon executives said the Intel money was used to spur the construction and equipping of its new 300mm memory wafer production fab in Dresden, Germany, which would increase output of all types of memory chips, including RDRAMs, to support all Intel microprocessors.

In other RDRAM developments, Samsung executives told the Intel Developer Forum this week in San Jose that a new four-memory-bank version now in development could reduce Direct Rambus chip costs by more than 20%.

When the four-bank chip enters full production in 2002, it could be close to parity with SDRAM prices, said Jon Kang, Samsung's senior vice president for memory product planning and applications engineering. The four-bank RDRAM depends on a new Intel chipset slated to be introduced next year, according to Kang.

Pete MacWilliams, Intel Fellow and a specialist on the company's memory road map, confirmed the new Intel four-bank chipset will debut early in 2002. It will support both a single-memory channel for mainstream PCs and a dual-memory channel used by the current Willamette Pentium 4, he said.

Intel has not yet selected a name for the new four-bank RDRAM chipset, according to MacWilliams.

Samsung's Mackowiak said the existing 32-bank Direct Rambus DRAMs were originally designed to meet the needs of server memories. When server OEMs opted to use SDRAM and DDR instead, the opportunity opened up to design a four-bank RDRAM for desktop PCs, he said.

Even before unveiling the new four-bank RDRAM, Samsung is steadily increasing its RDRAM production from 8 million 128Mbit-equivalent units a month this quarter to 11 million units a month in the second quarter, 13 million monthly in the third quarter, and up to 14 million per month in the fourth quarter.