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 : Intel Corporation (INTC) -- Ignore unavailable to you. Want to Upgrade?


To: John F. Dowd who wrote (89168)9/30/1999 4:56:00 PM
From: Richard Habib  Read Replies (2) | Respond to of 186894
 
Samsung to halt RDRAM production. Reliance reported to be developing DDR based mobo for Coppermine.

Samsung halts Rambus production after Intel delays Camino

By Will Wade
EE Times
(09/30/99, 3:19 p.m. EDT)

SANTA CLARA, Calif. — With another delay of a key Intel Corp. chip
set for enabling Direct Rambus DRAMs in PCs, Rambus systems have
missed the window for the Christmas selling season, analysts say,
potentially costing PC makers tens of millions, possibly hobbling the
upcoming launch of Intel's next version of the Pentium III and causing at
least one DRAM maker to halt production of Rambus parts. The delay
could also help give double-data-rate SDRAMs a stronger chance at
becoming a mainstream alternative for high-end PCs sold next year.

Companies at every level of the supply chain are reeling from the impact of
Intel's last-minute delay of the 820 chip set, code-named Camino.
Designed to link direct Rambus memory chips with a microprocessor, the
product has been seen as the critical last link needed to deploy
RDRAM-based PCs.

Intel had planned to officially introduce the Camino this week, and
numerous PC OEMs were poised to simultaneously unveil systems using
RDRAM technology. Instead, Intel confirmed that the chip set has failed
validation tests, postponing the launch indefinitely. Instead of enjoying a
long-anticipated ramp for RDRAM technology, the unexpected turn of
events means PC makers, memory vendors and other players in the market
must re-evaluate their Rambus strategies. Adding insult to injury, nobody
knows whether the products that have already been produced can be used,
or if they are destined for the scrap heap.

Faced with an uncertain future for RDRAM sales and rising prices for
SDRAM products, memory vendors are starting to shift their production
mix. "There will be no more wafer starts for RDRAM until we can better
understand how long it will take to resolve the Camino situation," said Avo
Kanadjian, vice president for memory marketing for Samsung
Semiconductor Inc. (San Jose, Calif.). "Any capacity that can be freed up
will be reassigned to 128-Mbit or 256-Mbit SDRAM products. We will
require some convincing before we restart any RDRAM production."

RDRAM and SDRAM technologies use entirely different mask sets, and
shifting back and forth is no small task, although Samsung's Fab 9 was
designed specifically to manufacture either type of memory. Kanadjian said
the company's current RDRAM inventory and the work now in progress
totals some 100,000 Rambus chips. If Intel cannot identify the bugs in its
chip set soon, he said, that should be enough to supply PC OEMs through
the end of the year. And by shifting to SDRAM, he said the company could
pump an additional million SDRAM chips into the channel by year's end.

With SDRAM prices on the rise, the glitch could be an opportunity for
memory companies to adjust their mix and increase their revenue in the
year's final quarter. Other memory vendors are reacting to the same issues,
and Hitachi Ltd. said it they will hold off on ramping RDRAM production
pending more solid information about Camino's launch.

PC manufacturers could face even bigger problems since they must also
delay the introduction of RDRAM-based systems. "The PC companies are
about to go into their biggest season without any whizzy systems to put on
the shelves," said Sherry Garber, memory analyst for Semico Research
Corp. (Phoenix). "Instead, they have nothing."

One analyst estimated that if OEMs have to scrap their existing Rambus
motherboards it could cost them $10 million based on an estimated
100,000 boards. But another source said as many as 500,000 Rambus
boards may have already been manufactured based on orders of key
components.

Ron Bechtold, vice president of sales and marketing for Hitachi
Semiconductor (America) Inc. (San Jose, Calif.), agreed that the delays
make it likely Rambus will miss the critical holiday window. "You probably
won't get a Rambus computer in your stocking this Christmas," he said.

A spokesman for Dell Computer Corp. (Round Rock, Texas) confirmed
that the company will not produce or ship any boxes with RDRAM until
Intel can correct the Camino bugs.

At this point, Dell and other PC vendors are waiting for Intel to say
whether potential workarounds exist. In the meantime, those designs sit
unused. "It's always a problem when you plan to be selling something, and
it is delayed," the spokesman said. Unlike other vendors, Dell's
build-to-order model means they have not yet begun producing the boxes
for shipment and will continue to produce SDRAM-based machines only.

If the PC companies have to scrap their RDRAM-motherboards, it will be
several months before the technology can debut, said Peter Glaskowsky,
senior analyst with MicroDesign Resources (Sunnyvale, Calif.). And all of
that is contingent on Intel's finding the cause of the problem.

"We are continuing to work on identifying the cause of the errors, but at
present there is no new launch date," said an Intel spokesman. "We will
introduce the Camino when it is ready for high-volume production."

The problem seems to lie in the way the Rambus memory slots interface
with the rest of the motherboard, Glaskowsky said. The Rambus design
allows for three memory slots containing up to 32 chips each. A typical
OEM strategy has been to put all three slots on the motherboard but fill
only one, leaving space for expansion.

Slipshod slot

But the third slot in some cases behaves unreliably. "Intel basically had to
advise [PC OEMs] they can't use all three slots," he said. "It's my
understanding they can't ship motherboards they've already manufactured,"
said Glaskowsky. The problem occurs even if the third slot is empty,
although it is "probably worst in the more heavily loaded configurations," he
said.

Indeed, Intel's inability to identify the exact problem is causing stress in the
industry. "Intel has told us the delay is indefinite," Kanadjian said. "It is
preferable if they could give us some kind of schedule so we can plan our
own mix, but I think it's clear that they have already missed two delivery
dates and they want to be certain that the next one is 100 percent
achievable."

Intel had originally planned to roll out the Camino last spring, but said in
February that the schedule would be pushed out for three months. The
latest problem scrapped the second planned launch. "This certainly hurts
Intel's credibility," said analyst Garber. "Two delays is very serious."

While this delay is embarrassing, said Nathan Brookwood, principal analyst
for Insight 64 (Saratoga, Calif.), the real test will come later this month
when the company is expected to roll out the 0.18-micron version of its
Pentium III, code-named Coppermine. That chip will also integrate the
Level 2 cache, which generally brings about a 10 percent performance gain
over similar chips running at the same clock speed but with off-chip cache.

Brookwood said Coppermine is the chip most PC vendors expected to
really activate the Rambus rush. "If Intel can't correct the Camino situation
before the Coppermine debut, that will really throw a wrench into their
plans," he said. "[Intel] doesn't even know yet what the problem is. This is
an embarrassment to be sure."

The delay could give enough time for some alternative memories to gain a
stronger beachhead. Garber suggested that formats such as PC133 chips
or double-data-rate SDRAM might be able to capitalize on the Rambus
misstep, especially if RDRAM misses the entire fourth-quarter window.
"Shifting to 133-MHz SDRAM," she said, "could offer enough of a
performance boost to tide the industry over for a while," at which point it's
not clear that Rambus will be the only viable contender for next-generation
memory dominance.

One company suddenly in the driver' seat is Via Technologies Inc. The
Taiwanese chip-set vendor is now the only supplier of a chip set able to
support 133-MHz memory chips. In the wake of the Camino debacle, Via
may find its product in sudden demand. "This is going to have a strong
impact on our sales," said Dean Hays, director of marketing at Via's U.S.
subsidiary in Fremont, Calif. "We always say our success is based on our
own execution, but the delay of Intel's 820 chip set is all gravy for us."

Another chip-set maker, Reliance Computer Corp., is said to be preparing
a chip set that will support the Coppermine CPU and use DDR memories.

Jim Handy, memory analyst for Dataquest Inc. (San Jose), still expects to
see Rambus emerge in the long run as the market's dominant DRAM
format. "Intel is pushing the market in that direction, and that doesn't leave a
lot of choices for the rest of the industry," he said. "Alternative memory,
such as DDR, would need a lot more time than they will get from this delay
to really become established. Rambus has soundly missed the Christmas
market, but they will probably ramp next year."

Bechtold of Hitachi, which is not a strong proponent of Rambus, said the
latest delay may make Intel relax its corporate push for RDRAM
memories. "I think they will begin to let individual product groups have
more sway about choosing which memory type makes sense for their
designs," he said.

Though Rambus' stock value plunged some 30 percent on news of the
Camino bugs, executives at Rambus remain optimistic about the long-term
picture. "We still see it as inevitable that RDRAM will become the
dominant memory technology," said Dave Mooring, senior vice president
for the Mountain View, Calif.-based company.

Separately, Fujitsu Ltd. had decided even before the most recent Camino
delay not to enter the Rambus market, according to Keith Horn, director of
memory marketing for Fujitsu Microelectronics Inc. (San Jose). The
company had been sitting on the fence for some time, going so far as to
produce a prototype of a 72-Mbit RDRAM, while holding off on making
an investment in test and packaging infrastructure until the market situation
became clearer.

The internal debate ended recently as Fujitsu decided to put its efforts
behind its double-data-rate (DDR) and fast cycle DRAMs instead of
Rambus. "We're seeing a lot of interest in those products, and decided to
focus on those efforts," Horn said.



To: John F. Dowd who wrote (89168)9/30/1999 8:28:00 PM
From: Paul Engel  Read Replies (2) | Respond to of 186894
 
John - Re: "It amazes me how stupid the "big boys" "

I'll say.

With the recent panic caused by the Taiwan Earthquake and today's "potential" problems for Japans semiconductor manufacturers due to the nuclear release, I have yet to see ONE ANALYST praise Intel for making a strategic decision in 1974 - 25 years ago - to disperse its wafer fab manufacturing operations ACROSS the United States and more recently internationally.

With Fabs in Portland. Oregon, Phoenix, Albuquerque, Santa Clara, Ireland, Israel (and later, Texas) the geographical dispersion supplies some insulation to localized problems.

But the very few analysts understand this.

Paul