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: Joe Donato who wrote (25231)	7/19/1999 9:32:00 PM
From: KM	Respond to of 93625

Heeeere's Marcy! Intel's Eye Wanders Away From Rambus By Marcy Burstiner Staff Reporter 7/19/99 9:04 PM ET SAN FRANCISCO -- The lesson from Rambus (RMBS:Nasdaq) this month: Don't invest if you have a weak heart. Just when investors were starting to breathe easy after months of speculation that Intel (INTC:Nasdaq) might back away from Rambus' revolutionary memory design, Intel officials pop up to say they'll evaluate the PC-133 technology set up as an alternative to Rambus chip designs. That's the very technology that Intel has repeatedly said it had no intention of adopting. The result? Rambus shares dropped 12.7% to close at 98 5/8 Monday, when the Nasdaq Composite index dropped just 1.2%. "There is a lot of pure brutal trading around this stock one way or another," says Hambrecht & Quist analyst David Wehner, who upgraded Rambus to buy from outperform last Thursday. "I wouldn't be surprised if at some point Intel supports PC-133." H&Q has an underwriting relationship with Rambus. Intel spokesman Mike Sullivan told TheStreet.com Monday that Intel, on the request of makers of computers and memory chips, will evaluate a PC-133 synchronous DRAM chipset as a stopgap alternative to more expensive Rambus-based chipsets. Intel's chipsets -- devices that connect a computer's memory to the microprocessor -- are important because of its iron grip on the chip market. The OEMs fear there will not be enough Rambus-based chipsets on the market, Sullivan said. On Friday, Rambus hit an all-time high of 117 1/2 after surging 30% in two days, powered by strong third-quarter earnings and news that Rambus' planned rollout this quarter was on schedule. The news also coincided with introduction of a PC-133 chipset from Intel competitor Via Technologies of Taiwan. Intel executives have been asked repeatedly at investment and technology conferences if the company would support PC-133, but until now have denied the company would do so. "We have been saying that PC-133 doesn't really make sense," Sullivan said. PC-133 is seen by some in the industry as a stopgap measure until faster, non-Rambus memory chips can be developed. The memory industry has balked at adopting Rambus-based standards because companies licensing the technology must pay Rambus royalties that average 1.5% of the price of each chip sold. Sullivan said more definitive news will likely be released at Intel's three-day Developer's Forum in Palm Springs, slated to begin Aug. 31. Wehner downplayed the significance of the news. Many Rambus investors, he says, have expected to see PC-133 enter the market and Intel's possible adoption of it won't necessarily hurt any Rambus rollout. "I'm pretty comfortable with my numbers for Rambus," he says. "I think Rambus can chug along quite fine." At the same time, he says, investor reaction is difficult to predict on this stock. "It's hard to say what the individual investors are thinking," he says. "I don't know what the stock will do tomorrow." --------------------------------------------------------------------------------

To: Joe Donato who wrote (25231)	7/19/1999 10:25:00 PM
From: Carl R.	Read Replies (3) \| Respond to of 93625

I knew a guy once that had a special engineering simulation program that had only one loop that was about 16MB in size, and was executed many, many times. His is the only application I can think of that would benefit significantly from a faster CPU-memory interface. For other applications, where an 80% L1 cache hit rate, and a 90%+ L2 cache hit rate is the norm, I wouldn't expect to see more than a 3-5% improvement. From my experience improvements smaller than about 30% are not noticeable to users not holding stopwatches. How much more will users be willing to pay for a product that produces no noticeable benefit, especially in an era of super low-cost computers? Carl

To: Joe Donato who wrote (25231)	7/20/1999 1:27:00 AM
From: Tenchusatsu	Respond to of 93625

<What problem is DRDRAM solving that PC133 doesn't? and, to what extent is this problem pervasive in the PC business.> DRDRAM is providing a real memory bandwidth boost, compared to PC133 SDRAM which is only providing a small improvement. You are right. For most desktop applications, the L2 cache does an excellent job in working around the limitations of memory bandwidth. But that would beg the question of why we should even go to PC133, when PC100 is good enough for those desktop apps. Heck, Celeron still only requires PC66, even though at higher clock speeds, the slow SDRAM does take its toll. On the other hand, there are plenty of high-end applications that can truly benefit from the bigger bandwidth. Quite a few of the "Pentium III" applications that require SSE (speech recognition, for example) benefit tremendously from memory bandwidth. Same thing goes for graphics workstation stuff, like 3D animation, CAD work, and engineering modeling and analysis. Although I have not seen any real benchmarks, I will bet that those types of applications will see some immediate benefits come "Rambus Christmas" in September. (To be sure, I'll also bet that there will be many situations where DRDRAM provides no benefit whatsoever, at least for the near future.) Also, DRDRAM solves the problem of scalability. Even if DRDRAM is shown to be useless at this point, future processors will no doubt get faster and faster. In this case, Intel's push for DRDRAM is more of a forward-looking thing, to get the transition pains out of the way so that when faster processors and AGP cards are released, when the applications and the games demand even more computing horsepower, the memory will not be as severe of a bottleneck. For the long-term, DRDRAM is more sophisticated and more scalable than any SDRAM technology, whether its PC133 or DDR SDRAM. Of course, the counter-argument is that no one needs the power. That's the same song-and-dance I've heard over and over again. While people who buy sub-$500 PC's probably don't need the power right now, the professionals do. And those professionals are the ones who pay over $2000 per PC, and who won't mind paying the initial premium for DRDRAM. Furthermore, the general trend of technology is this: Whatever is cutting-edge and expensive today will become mainstream commodities tomorrow. And even for the sub-$500 PC, DRDRAM can serve in two ways in order to contribute to cheaper system costs. First, a DRDRAM controller can be integrated onto a processor die because of its low pin count. That not only saves on chipset costs, it also improves performance by reducing latency to memory. And second, as DRDRAM chips become denser and higher in capacity, you can go with fewer and fewer chips, even to the point where one or two DRDRAM chips soldered onto the motherboard can hold 64 MB of memory. With SDRAM, you can't go below eight chips without incurring a performance penalty. So as of now, the argument can be made that DRDRAM is a solution looking for a problem. However, my spin on things is that DRDRAM is a solution ready to handle the future. And that future will come before we even see it on the horizon. Tenchusatsu