Silicon Investor (SI) -- The First Internet Community

STOCKTALK

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: Dave B who wrote (36991)	2/11/2000 4:05:00 PM
From: Dave B	Read Replies (1) \| Respond to of 93625

Part 1 of 5 Okay, here it comes. I'm blasting out the notes that I took (9 pages). Some of these points are repetitive, or things that most of us know already, but some are not. Anyone who was there is welcome to disagree or correct my statements. I was scribbling furiously and may have missed/misunderstood/misinterpreted anything. I'm still amazed each year that the official business at these meetings is finished in ten minutes. We had an attempted disruption of the proceeding this year by several guys who tried to second Harmon's two motions (to close the meeting and something else) faster than the two women from Rambus who are planted in the audience each year to "so move" and "second" the motions. Fortunately, the two guys couldn't out-move the two women and everything proceeded as planned <G>. Overall impressions - as with lunch, the meetings are getting larger (Tish apparently missed the first two). The first year I think there were 3 of us that weren't from the company or auditors. Last year maybe a dozenish (or a little over). This year probably a little more than two dozen (it's getting hard to tell exactly who's from the auditors and Rambus and who's not). - The presentation after the basic business followed the same format as previous years, though the content has changed. - As I've mentioned already, I saw a little more fire from Tate, especially when discussing the lawsuit, than I've seen in the past. I was very surprised that they are not interested in licensing Hitachi, but would rather shut down Hitachi's SDRAM, DDR DRAM, and SH microprocessor products. - Harmon presented first (through the end of Phase I below, then Tate took over) Market Share Forecast Dataquest and Instat both forecast the following market shares for RDRAM (of the DRAM market) 1999 tiny 2000 10% 2001 30% 2002 50% 2003 60% RDRAM Systems in Production There are currently 18 RDRAM-based systems available for purchase from Dell, HP, Acer, IBM, Compaq, Micron, and Gateway. The PS2 Harmon covered again the fact that the PS 2 would have 2 128Mb RDRAM chips. He said that showing RDRAM working in a low-cost, low-end platform like this would have an effect on the consumer market. A little later in his presentation, he claimed that if the PS2 had used SDRAM or DDR, the number of pins on the CPU for memory access would have gone from the ~140 for 2 RDRAM channels to ~200 to ~400, requiring 16 SDRAM chips or 8 DDR chips. They estimate that this would have added $30 to $50 to the cost of making the PS2.

To: Dave B who wrote (36991)	2/11/2000 4:05:00 PM
From: Dave B	Respond to of 93625

Part 2 of 5 Competition PC133 - is mainstream now. Systems will switch to RDRAM as the demands on the system increase. These demands will be driven by applications like multimedia and games. If you want to use your PC as a typewriter, you don't need Rambus (his words). DDR - is evolutionary from SDRAM (though later he said it was not evolutionary because you couldn't use SDRAMs in a DDR motherboard). In the interim this will be used in servers, which he called a niche market. As for DDR in PCs, RDRAM has a faster per pin rate and is scaleable since you can add RDRAM channels more easily (no change to the story). In Value PCs, he called it a myth that you needed RDRAM only for performance and used the PS2 as the example. You can get the full bandwidth of the technology from a single chip, which will be less expensive than 8 or 16 chips. He also showed a line drawing of a TI DDR DIMM which had 11 additional support chips on the DIMM for buffering, latching, et cetera. Lots of chips! With RDRAM in production now, with over 50 controller design wins, and with DDR at least a year behind RDRAM, Harmon called it ludicrous that anyone would forecast that DDR would outship RDRAM in 2000 (take THAT, Sherry!). Advanced DRAM Technology Consortium - he said that this is an admission that the SDRAM/DDR technology is at the end of its life. The consortium is targeting 2003 as the date to release their new technology. That gives RDRAM four years to penetrate the market before they even show up. And once they do show up, they are going to have to be a leap ahead of RDRAM to convince everyone to put in place a new infrastructure to support the technology. Also, if they try to implement any technology that's packet-based, they'll probably run into Rambus patents. Finally, Harmon pointed out that consortiums have not worked in the past very well. Tate, in the Q&A session, supported that statement be saying that the biggest problem with consortiums is goal congruence. Rambus has been talking to the members of the ADT and they found that the manufacturers have different objectives for the consortium. This will make it hard to accomplish anything. He pointed out that one of the members even said that they expect it to take a year just to get the group infrastructure defined and in place. Patents Harmon presented the following table for patents filed and issued (I didn't get the 1995-1997 numbers for patents filed if anyone else has them)? Filed Issued 1995 17 1996 27 1997 35 1998 120 50 1998 160 70 Licensees Harmon showed a slide with a bunch of names - 32 semiconductor manufacturers have now licensed RDRAM technology, along with scads of infrastructure vendors (testers, connectors, RIMMs, clock chips, etc.)

To: Dave B who wrote (36991)	2/11/2000 4:06:00 PM
From: Dave B	Respond to of 93625

Part 3 of 5 Testers Gary wanted to debunk the myth that there were no demand for testers. He said that dozens and dozens of testers have been ordered from and shipped by Schlumberger, HP, and various Japanese companies. He said that Rambus was not aware of any orders for Teradyne testers. [Note: didn't Teradyne already admit that it was there own problem?] Rambus - Phase I Harmon summarized Phase I of Rambus (1990-1999) as: - IP Development - The application of their technology to the memory interface - $1.25B of Rambus-based products have shipped - By the end of the year Rambus expects royalties to make up 50% of their revenues (up from about 20% now) Rambus - Phase II Tate took over next. He described Phase II as: - Solidify position in PC and consumer digital video markets - Push Rambus standard by advancing their technology lead - Apply the standard to new markets (e.g. communications) - License their IP for non-compatible uses (e.g. SDRAM, DDR) - Leverage the IP business model RDRAM Cost Differential Yes, RDRAM is more expensive. Tate says that it's currently 40% more expensive to produce RDRAM based on the feedback they're getting from the manufacturers. Pricing, on the other hand, is completely supply/demand driven. The following table shows the relative breakout of the incremental costs of RDRAM over SDRAM in 2000, and the expected improvements for 2001 (as an example it costs twice as much, or 100% more, to package the RDRAM today as it costs to package SDRAM): 2000 2001 Die size 25% 10% Test 55% 1-2% Package 100% 10% Total 40% 10%

To: Dave B who wrote (36991)	2/11/2000 4:06:00 PM
From: Dave B	Respond to of 93625

Part 4 of 5 Rambus Technology Future As has been stated before, Rambus expects to double the data rate per pin from the 800 megabits per second (Mbs) currently to 1.6Mbs in 2000. Likewise the module performance will quadruple from 1.6 gigbytes per second (GBs) to 6.4GBs. They will announce this formally in June and provide a technology paper to describe it. They are close to the first customer commitment for this technology. The Communications Market Tate showed a graph that I can't recreate that had the performance level of communications products (from .6Gbs to 40Gbs) up the left side and showed "Integration/More services" along the bottom, listing (from left to right) "Buffer", "Look-up Tables", and "QoS" (quality of service). A line on the chart ran from the upper left corner to the lower right, implying that Buffer, for example, was already at the 10Gbs to 40Gbs range, while QoS was down in the .6Gbs to 1Gbs. He said they probably have about 10 design wins for leading edge technologies which won't represent much volume now, but as the high-end technology becomes the mid-range technology, the volumes will grow. Available Rambus Market In 2002/2003, Tate says the available market for Rambus technologies is roughly $100B in memory chips and logic chips to interface with those memory chips. Also, there's another $150B in Other Chip-to-Chip Communications products that bring the total market to $250B in 2002. [Don't forget to take the royalty % of this, then reduce it further by a market penetration factor]. Rambus IP Rambus' fundamental patents on high bandwidth memory subsystems date from 1990. Rambus is willing to license the technology for non-RDRAM use at reasonable rates. The initial focus of Rambus efforts to license the non-RDRAM technologies will be on companies not helping to grow the RDRAM market. THE LAWSUIT Rambus and Hitachi signed an agreement in 1992. To date, Hitachi has not produced a single IC supporting Rambus technology, nor have they met a single milestone. They chose to file the lawsuit in Delaware as things move more quickly there, and they expect that if it goes to a jury trial, it will take 12-18 months. The objective of the suit is to gain an injunction against the import, sale, manufacture, and use of Hitachi products that use the Rambus technologies. Specifically, these include Hitachi SDRAM, DDR DRAM, and the SH microprocessor line. They are not planning to let Hitachi just do what they would have done if they had licensed the product. In other words, Rambus is not going to spend $M just to get them to pay royalties - they expect to stop Hitachi from producing these products.

To: Dave B who wrote (36991)	2/11/2000 4:06:00 PM
From: Dave B	Read Replies (8) \| Respond to of 93625

Part 5 of 5 Q&A The following were some of the questions and answers that weren't covered in the previous messages. Is embedded DRAM a competitor? Not really. Right now you can't embed much. Three years ago, the forecasts said that embedded memory would be much bigger, but it's not. Beyond 2010, we don't know. What was the problem with the 820? Think of Intel as the general contractor who says "do this" and we do it, but we don't know everything that's going on. It was a joint problem (we're not blaming Intel), and given the FDIV problem that Intel ran into, it wasn't acceptable to release the product as it was. The problem has been nailed down. [Note: I have no idea what this means, either.] How are royalties structured and what are the rates? There are all sorts of variations with caps, minimums, step down rates, etc. so there's no standard deal. Intel obviously got a really good deal. For RDRAMs it's up to 2.5% and for logic chips 3% to 5%. Why the lawsuit? We tried to work with Hitachi. The missed all their contractual obligations. One effect is that all the other licensees are being nicer. Also, we've been trying to higher additional lawyers to look at the patents and are having trouble finding firms that aren't already looking into the issue. Which tells us that a lot of lawyers are studying our patents. Why is there such resistance from the DRAM manufacturers? It's a competitive game, and they don't want Samsung to win. It's more important that the PC company executives are lined up, since the DRAM guys will make whatever gets ordered. The PC execs are saying that RDRAM is a major, and growing, part of our product mix. The interests of the DRAM manufacturers and systems companies are in conflict - the system people want fast, low-priced, commodity-like memory, whereas the DRAM manufacturers want a differentiated, fractionalized market. [Note: he added later, in the stand-up session, that the public perception lags the internal perception by about 6 months. Once the 820 started shipping and the orders from the system manufacturers started coming in, the DRAM manufacturers began taking those orders and starting up the lines, but the public doesn't see that yet.] Is there a microprocessor speed where RDRAM is the only answer? It all depends on the software. What about RDRAM in servers? Most Sun system still ship with EDO DRAM. The server manufacturers tend to be a couple of years behind as they tend to be cautious and they have slightly different needs. Is the Toshiba redesign of the DRAM core (which gets the die size penalty down to the projected 10% now) proprietary or available to all? If it's a redesign of the core, it's probably proprietary, but that doesn't mean the other manufacturers can't figure out how to do it as well.