Intel can dance, no doubt
How the semi firm waltzed through the memory wars
By Bill Teel: Friday 25 April 2003, 19:24
dg> Formerly?? .... Bill Teel Director of Marketing Fredhager.com
dg> Sorry! Everything is written with a slant to it!
THERE IS A THEORY that Intel still doesn't want to back away from Rambus' DRAM memory designs. And the company has no intention of doing so, despite the lack of any chipsets on their current roadmap with RDRAM, as well as its publicly agnostic position on memory.
It's believed that genuine market dynamics killed RDRAM's chances. Yet, from the evidence that is surfacing in the FTC's anti-trust case brought against the intellectual property firm, it seems, perhaps, a whole other story played out, and there was a handful of DRAM players who forced Intel to abandon the technology by collectively boycotting RDRAM.
From evidence in the FTC case, it appears that once certain DRAM members set their sites on derailing RDRAM, there was little stopping them from following through on a "united strategy"¹ to stop Intel from succeeding with a new RDRAM standard.
Instead of strong arming its DRAM suppliers into producing RDRAM (although RDRAM made the most sense from a long-term performance standpoint), Intel was simply smart enough to step back while the mayhem ensured, write-off the substantial investments it had made in many of the manufacturers to ramp up RDRAM, then pick up the pieces later and carry on.
When the industry started expressing, through harsh criticisms of Rambus and RDRAM in the press, as well as inactions in manufacturing, the united resistance to stop RDRAM became, in hindsight, quite obvious. With the deck stacked against it, RDRAM didn't have a chance.² Considering there is a strong possibility that Rambus will be able to prove willful infringement of DDR, it seems the industry not only kneecapped RDRAM's adoption, it lifted parts from Rambus' design to ensure their DDR DRAM could appear relatively competitive.
If things went Intel's way, and if the majority of today's PCs contained RDRAM memory, and not DDR, think of the outcry from the memory firms now. In that hypothetical all-RDRAM world, Intel would likely have a stack of anti-trust claims against them piled high on their desks, and they would be spending plenty of time, and dollars, in the courts fighting to convince judges that RDRAM was, "really for the sake of performance."
In today's real world, it's debatable whether or not the memory firms would be making more money with RDRAM than DDR, despite the manufacturing transition costs to RDRAM, and the royalty. (Rambus fetches approximately a 1.5% royalty for each RDRAM chip the memory manufacturers sell.) On one hand, the transition from SDRAM to RDRAM would have justified higher prices from everyone – more so than the supposedly easy transition to DDR. Samsung did it, and the company attributes their profitable DRAM business in 2001 largely to RDRAM.
On the other hand, from the apparent actions of the more contentious memory firms, it's possible they would be blaming Intel for forcing them into RDRAM, and subsequently running out of money because they couldn't transition fast enough, or make a margin.
As the downturn in the market proceeded, Intel's decision to back off and let the memory players upgrade at their own pace, although pushed by the performance of RDRAM, seemed to be the easiest and safest solution. While Rambus took most of the heat for the missteps in the transition to RDRAM, investors began to wonder if Intel had made a big mistake by going all RDRAM, especially considering AMD's revival at the time. Supporting DDR would remove the uncertainty of RDRAM in a time when investors were mired in fear of their technology investments.
Many thought Intel was foolish for letting RDRAM go as easily as they seemed to, but the evidence produced in the FTC case against Rambus is telling, and may change a few minds. Since it appears there was no stopping the DRAM manufacturers from killing RDRAM, Intel was better off, “letting the market decide.” By choice or not, Intel allowed the market to choose DDR as the new mainstream memory; the only problem is, the industry may not have reached that conclusion fairly, and many have practically driven themselves out of business in their campaign to secure DDR's success.
That Was Then, This Is Now
Selfless, Stupid or Savvy?
Recently, SiS officially announced that it had signed a long-term licensing agreement with Intel that will allow it to produce chipsets for the new 800-MHz system bus. Intel and Via have also recently settled their various patent lawsuits, and as a result, Via has a similar licence.
So, today, Intel has opened their doors to the chipset makers on the rise, and given SiS the platform by which it may legally blow the doors off of any Intel chipset in existence.
Graciously granting SiS the licence to the 800-MHz bus a quarter away from SiS's debut of its quad-channel RDRAM chipset is one heck of a way to show that the market may want something other than DDR.
After all, if people appreciate SiS's 659 at all like they did Via's DDR DRAM P4 chipset, why wouldn't Intel go ahead and make an SiS-like chip for itself, too? Hey, if the market took RDRAM off of Intel's map, they might as well put it back on there. It's anti-trust free, and Intel might be happy to play catch-up in this scenario.
In Via's case, two short years ago the company was a big thorn in Intel's side, and a reason for RDRAM's failure. In 2001, Via started making a 400-MHz P4 chipset compliant with DDR DRAM. Intel only wanted to match the P4 with RDRAM, and Via's entry, while without a licence, showcased the gap in pricing, and put the pressure on Intel.
Ultimately, Via's scheme helped to keep RDRAM pricing in the stratosphere, and a seemingly unnecessary upgrade. Lucky for Via, but to just about everyone else's misfortune, the market crumbled, and with it the demand for PC's shrank, making DDR DRAM an easy alternative that saved a few bucks. (At the time, memory manufacturers were selling DDR at SDRAM prices.) Before Intel could do anything about it, Via had grabbed too much share, and a DDR DRAM P4 chipset made its way onto Intel's roadmap.
From A Handout To A Leg Up
While the SiS and Via agreements both signify a turn for Intel, licensing doesn't appear to be the only place the company seems to be sharing the love. The rumours of Intel helping out Elpida with its fab expansion intentions continue to circulate, and there couldn't be a friendlier recipient. Elpida, while the fifth largest DRAM manufacturer, is right up there with Toshiba as one of the top three producers of RDRAM.
Some consider Intel's licence to SiS an odd event, considering it will relinquish Intel's performance crown. But, taking into account the work Intel has to do to keep its DDR chipsets running smoothly, it might as well unofficially outsource some of their R&D to eager engineers at Rambus, SiS and Elpida. In Rambus' case, if it feels it has to prove something, the new SiS quad channel chipset could be a great way to do it.
Coincidentally, or not, Elpida is an early licensee of Rambus' Yellowstone and Redwood technologies, and that could very well come in handy if Microsoft wants to use either, or both, with the help from Intel, for the next X-Box game machine.
Samsung, the top DRAM manufacturer of all DRAM flavors, including RDRAM, has not yet taken a licence on Yellowstone, or Redwood. But then, no company, except perhaps Intel, knows Rambus' technology as well as Samsung, so pursuing Yellowstone and Redwood seems fitting for Samsung.
The press releases from Rambus and their partners upon the licensing of Yellowstone mostly discuss the applicability and cost savings of implementing Yellowstone and Redwood into games, consumer electronic devices and communication systems. But make no mistake; Yellowstone is an ideal next generation PC DRAM solution.
Wait And See…
There are possibly more synergies, and safe havens, for Intel's actions as of late than what meets the eye. To some, Intel needs Rambus as much as any partner to stay competitive, and Yellowstone and Redwood make the most sense long-term. Considering Intel's comprehensive patent licensing agreement with Rambus, it seems unlikely Intel is only utilising the portfolio for their communications efforts – especially considering Intel bought into Rambus for its DRAM expertise in the first place.
In less than a week, the FTC will kick off its anti-trust trial against Rambus and decide if the company tricked the industry into using its "to-be" patented SDRAM and DDR technologies.
Whether or not the evidence that supports an alternative anti-trust view will be allowed at trial is in question, and it remains to be seen what Rambus does with the information it's uncovered.
In light of the compelling evidence produced in the FTC case, as well as the Department of Justice's ongoing criminal investigation into a possible price-fixing scheme – one that was possibly orchestrated by some of the DRAM manufacturers (not Rambus) – a different take on the events of the last few years may be worth further attention.
¹ The minutes of a December 3, 1996 meeting of membership-restricted manufacturer consortium called, "SyncLink Consortium" state that "many suppliers are paranoid over the prospect of a single customer, e.g., Intel having control of market. We can't resist such a possibility individually. We need some united strategy." In a later meeting, manufacturer representatives acknowledged that, "Intel won't change course unless Rambus fails." – From "Memorandum By Rambus Inc. In Response To Motion By Department of Justice To Limit Discovery Relating To The DRAM Grand Jury." Page 13. here.
² Among the messages delivered by Mr. McComas to DRAM manufacturers at the April 1998 seminar was a prediction that Intel was likely to try to force manufacturers to bear whatever higher costs might be involved in Rambus DRAM production, so that the Rambus DRAM is a "guaranteed bad bet for margin enhancement." (Id., ex. K.) Possible strategies to avoid this "bad bet" included "resist popular deployment of" Rambus DRAM. Id. One suggested way to accomplish this was by keeping Rambus DRAM production low: "tape out but do not fully productize or cost reduce" Rambus DRAMs.
At the June 1998 meeting of DRAM manufacturer executives, (Bert) McComas suggested he receive the manufacturers' Rambus DRAM production forecasts in order to create, and circulate, a combined industry forecast. (Id., ex. M.) As he explained in an August 1998 e-mail to a Hynix executive, this service would be useful because "during the critical production ramp-up phase of Direct Rambus, DRAM vendors will need a constant flow of information to help make wise decisions and to walk the fine line between a pleasant shortage and a disastrous oversupply" of Rambus DRAMs.
The Hynix executive who received McComas' e-mail conceded in his deposition that an "oversupply" of Rambus DRAMs would have been "disastrous" because the price would have gone "way down." (Id., ex. O.) He also agreed that Intel "couldn't start the ramp-up" of Intel products that incorporated Rambus DRAMs unless the price of Rambus DRAMs did come "way down," or at least came "very close to the industry standard." From "Memorandum By Rambus Inc. In Response To Motion By Department of Justice To Limit Discovery Relating To The DRAM Grand Jury." Page 13-15. See here.
Bill Teel is the Editor of Hedge Fund Confidential. He owns Rambus shares.
Thanks to REH Message 18885102 |
