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: Dan3 who wrote (26650)	8/8/1999 10:26:00 AM
From: unclewest	Read Replies (2) \| Respond to of 93625

Dell's paper seems to be making the same point that I was in its discussion of the future - that rambus's best bet is to drop in price and be used to save money through its low pin count and mboard trace count. in a post yesterday, you suggested that a way to that end would be for rambus to reduce their royalty. you further suggested the royalty to be 5%. rambus cfo said the pc dram royalty is 1-3%, average estimated to be 1.7%. royalties on other products go up to 5%. this is 1.7% of the memory not of the total pc cost. when we consider 145 million pc's estimated for next year, it amounts to a potentially huge royalty revenue base for rambus. when you look at the royalty per pc, it is a rather insignificant amount of total pc cost. the overall costs of rdram will come down imo when all dram mfrs are on stream in full production fighting for business. intel's timetable projects us completing that transition in less than two years. i predict the volume increases from rambus' increased market penetration and projected overall dram market growth will overcome the price reductions in the future allowing for continued rambus revenue growth. of course by then, rambus will be an enormous success with boatloads of cash to pursue other endeavors. did you know that rambus believes they can continue to increase speed by 100-200mhz annually? rambus 1000 rambus 1500 rambus 2000 oh no! we all have to upgrade again. this very recent statement from samsung certainly is in opposition to several of your comments. as a long, i find it quite comforting. DRAM performance is measured with two metrics: bandwidth and latency. Surprisingly, one type of DRAM delivers the highest performance in both areas - the Rambus© DRAM. It is widely recognized that with the availability of the 1.6 gigabyte per second Rambus DRAM (RDRAM©), sustainable system bandwidth has jumped a factor of 10 over SDRAM. What hasn't been as apparent is that RDRAM latency has also been improved relative to SDRAM. What may be even more surprising is that 133 MHz SDRAM latency is worse than PC100 SDRAM unclewest again i would like to point out that this thread continues to concentrate on pc sales. that is only 35% of rmbs total revenue base. 35% pc's 15% servers 50% other

To: Dan3 who wrote (26650)	8/9/1999 2:21:00 AM
From: J_W	Respond to of 93625

Dan, Wow, I spend the entire day at the hydroplane races (Seattle Seafair) and I get back to find 40 new posts, many referencing new information I hadn't seen before. Talk about information overload. <gg> I appreciate the fact that you took the time to fully read Dell's white paper on Rambus. Not everyone that should actually does this. Makes for a much better discussion for everyone. I consider these and other white papers to be very important as the companies writing them have much at stake. Their future success or failure is on the line. Contrast this with web sites that publish test results, where they risk relatively little in comparison and you can see why I take this position. If you look at the paper carefully, you'll see the implication that a substantially faster memory bandwidth, especially one that can stream data at those faster rates instead of bursting it, yields a performance increase. Between the lines, it seems to say that this will more than make up for a slightly higher latency. I would have to agree with your statement. Intel has also emphasized the bandwidth over latency argument. What Dell and Intel see are ever increasing bandwidth demands on memory, not only from the CPU, but from the AGP and PCI buses as well. A cache miss while memory is responding to an AGP memory request can get very expensive. This makes any small difference in latency between RDRAM and SDRAM a lessor issue. The memory subsystem must have enough overhead and capabilities to fully service all of it's masters. Intel, unlike Wall Street which has a very short term outlook, must look at system design issues in the long term. To do this they had to take a very educated guess as to what software and applications will require of their processors and chipsets years down the road (Dell also did the same thing in their white paper). What Intel sees is that while PC133, DDR may provide a short term solution, these technologies do not provide them with a long term solution. They are very concerned about the ability to scale upwards. Neither PC133 nor DDR use high quality transmission line techniques. As speeds increase, the need to do so will also increase. Since Rambus has patented their techniques, finding an alternative method not subject to Rambus patent infringement, may prove to be difficult. Add to that Intel's desire to incorporate a memory controller onto the processor chip, Rambus with it's low pincount becomes very desirable. They see the need to start the transition somewhere, sometime. If not now, then when? Rambus 300/600 has the streaming data advantage and a bandwidth that is 50% higher than PC100 - yet 300/600 has been rejected as too slow to compete with PC100 (it's as fast as, but more expensive than PC100). If the problem isn't latency, what is the problem? That's a nice question, but a question that no one can really give a final answer to, particularly when RDRAM is just now ramping up this quarter. Until the IDF at the end of this month, when Intel will release Camino specifications and test results, the best anyone can do is speculate based on tests using beta chipsets. And I believe those having beta chipsets are also under non-disclosure agreements. This has certainly led to an environment where there is much fiction to go with the facts. I will wait till the end of the month before I try to draw any conclusions. But I can hardly see where the question you raise will have any bearing on the long term acceptance of Rambus or not. If those 300/600 RDRAM chips end up in games or graphic boards, great. Regards, Jim