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: Zeev Hed who wrote (26633)	8/7/1999 10:51:00 PM
From: pompsander	Read Replies (2) \| Respond to of 93625

Zeev, I agree completely... I am puzzled by some on the thread who state that because the dominance of Rambus technology is not clearly evident on today's systems with today's application software, that there is no need to "create" Rambus; its technology is not needed. PC 100, PC 133, DDRDRAM are all sufficient. But that kind of thinking would keep us in the stone age. I would bet my house that in three years time we will all be taking streaming video for granted, and 3-d imaging and voice recognition have numerous business applications. The systems to create and support these multi-billion dollar concepts will be built when the processors, memory interface and application challenges are overcome. If Intel sees these apps in its crystal ball, it will build the processors and chipsets to drive them. That is how Intel will make its money. Rambus, from virtually everything I have read or heard, is the only meaningful solution to what Intel needs in three years to keep its traditional earnings stream rolling. So, I am long and holding on.

To: Zeev Hed who wrote (26633)	8/7/1999 11:21:00 PM
From: Dan3	Read Replies (2) \| Respond to of 93625

Re: without a high speed memory This may sound backwards, but I think that the beauty of rambus isn't that it's fast, but that it's cheap. Rambus seems to have been conceived around the time that a difficult move from 66MHZ to 100MHZ was going on, and 100 looked like the end of the line for capacitor memory. The pentium's bus overcame the lack of progress in memory speed increases by doubling the width of the bus from 32 to 64 bits - but the next step, to 128 bits would result in expensive and difficult to design mother boards and high pin counts. So rambus built a clever arrangement of 128 bit wide 100MHZ cells, then integrated a high speed control system to serialize the data over a high speed 16 bit bus. Not only is the 128 bit memory now easy to place on the motherboard, but rambus also interleaved the 100 MHZ cells in the same way DIMMS are interleaved on servers to provide for high speed streaming data rates. The result is excellent performance, and lower pin and trace counts. Great, so far. Trouble is, as clever as this design is, it increases the die size around 40%, and the chip structures that drive the data serialization to 300-400MHZ and transfer data at both edges of the clock hasn't been easy to produce. There also is a big lag before each set of transfers can begin. It takes 40 to 50 ns before any data starts to come from the chip - and that's on the 800MHZ (400MHZ double data rate) parts. I don't understand why it takes so long for a part running at 400MHZ internally, (even if the actual cells are at 100MHZ - that's still only 10ns) but it does. Now that wouldn't matter if capacitor ram technology had remained stuck at 100MHZ, but it hasn't. Instead, the same process that can produce the 800MHZ parts is now producing 166MHZ SDRAM. And some of those designs are being speced out at single cas latencies. (eg. 1 x 6ns instead of 2 or 3 times 6 ns) and 133 mhz SDRAM is being produced that transfers data at both edges of the clock just like rambus. For most mainstream software, latency is as important, if not more important than streaming data rates in determining performance. And standard SDRAM can be interleaved just like rambus is by using more modules, if the expense can be justified (as in a server). But DDR and 16 byte bursts (instead of 8 or fewer) will probably provide sufficient streaming data rates without resorting to interleaving. The net result is performance that is as good as, or better, possibly noticeably better, than rambus. And these technologies don't add significantly to the cost of the memory module. And you do pay more for rambus - first you have to buy more silicon for a given memory capacity, then there are royalties on top of that added expense. So why do I say that the beauty of rambus is not performance but price? It's because I think that the price of the silicon will come down rapidly over the next few years, while the cost of higher pin counts on chipsets and cpus, and the cost of motherboard cm2 will stay the same or rise. So ultimately, rambus could become a winner for the great mass of entry level computers, as well as cost sensitive devices such as video games, digital television, etc. But it'll only happen when the price comes way down, and that won't happen right away. Good luck to you, as I said before, market factors (eg. Intel regains control of the high end and won't support anything else) can easily overwhelm these suppositions, whether they are right or wrong. Dan

To: Zeev Hed who wrote (26633)	8/8/1999 6:31:00 AM
From: unclewest	Respond to of 93625

Long term, INTC will keep increasing the power and speed of its processors to keep high margin products "on the shelf", but without a high speed memory what good will it do? Thus in order for INTC to keep the performance peak at the processor and not memory, it must get RMBS off and out zeev, i agree. the oem's and software guys want this speed too. much of the world remains computer illiterate so the market will remain large for a long time. the part of the world that presently owns computers can also afford upgrades. upgrading is a huge part of the computer industry no need to give that up yet. the industry knows why we upgrade. they are determined to keep feeding us what we want. we upgrade for only one single reason...performance. performance comes in many forms. remember muscle cars in the 60's and 70's? comparing the performance of rambus using today's word processors is like putting a new, fin nor, 3 bearing, spin casting reel on an old heavy fiberglass rod. the results just won't demonstrate the full potential of the reel. put it on a sleek new graphite rod and you are now ready to play. heck, i cannot begin to tap the speed of the computer i have now. of course, i said the same thing with the used ibm i bought about 12 years ago...said it again with my old dell 486...and again with my p90 laptop...now, i know my 450 will be slow soon too... rambus is not a product for our present applications. comparing rambus performance with an office suite does not make sense. rambus is the future...but this future is almost here...just months away. this future includes video and audio apps most of us haven't dreamed of yet. it includes voice commanded pc's that work. it includes intense graphics and photo apps, not just for the pro's but for all of us. it includes game improvements our kids will love. think about all of the products that sony and toshiba make that use memory. surf their websites. there are many, many...many.. many reasons these two companies spent $1 billion dollars this year building a rambus rdram plant. think about matsushita and siemens and all of the memory using products they make. look at their statements of committment to rambus. go to the rambus website and look at the announced design wins. i have posted links to sites that point to rambus in a wide variety of consumer products other than computers.i am convinced that there are dozens of new rambus design wins waiting to be announced. think about ibm's june statement that they will produce rambus rdram. look at dell's recent statements. think about the dram manufacturers who have committed to rambus...fujitsu, hitachi, hyundai, lg semi, micron, mitsubishi, nec, samsung, toshiba, infineon and ibm. the shorts try to feed us a tom's hardware and mccomas diet. written and rewritten daily. sorry, it just doesn't compute. intel is not increasing processor and memory speed so microsoft word will run faster. intel is increasing speed so we will be able to enjoy the next generation of apps. these apps are not mainstream yet because our puters can't handle them yet. the circle between hardware and software development remains intact. unclewest

To: Zeev Hed who wrote (26633)	8/8/1999 5:16:00 PM
From: Jdaasoc	Read Replies (1) \| Respond to of 93625

Zeev: <<it must get RMBS off and out, better sooner than latter, even if it means a bad quarter here and there, IMHO.>> If this is any indication, Kingston is either worried why you would want to buy 100 RIMM's or is able to deliver. Checked Ingram Micro and Dell's Gibabuys online web sites and Kingston's RDRAM is not stocked SKU at this time at either source. Searched on "KRB" as manufacturers part number. However, once I see anyone stocking RDRAMs, I will comit my final aliquot of cash to buy RMBS stock. The long awaited delivery of RDRAM enabled computers will be very shortly there after. 206.135.146.107 Kingston Direct Rambus Engineering Samples Kingston offers the following 128Mb Engineering samples: Kingston P/N Description Device Count KRB600X18-8/128 128Mb, 18-Bit w/ECC, 600MHz 144Mb x 4 pcs KRB700X18-8/128 128Mb, 18-Bit w/ECC, 700MHz 144Mb x 4 pcs KRB800X18-8/128 128Mb, 18-Bit w/ECC, 800MHz 144Mb x 4 pcs KRB600X16-8/128 128Mb, 16-Bit, 600MHz 128Mb x 4 pcs KRB700X16-8/128 128Mb, 16-Bit, 700MHz 128Mb x 4 pcs KRB800X16-8/128 128Mb, 16-Bit, 800MHz 128Mb x 4 pcs Kingston has created detailed product specification data sheets for these modules. These files are in Adobe Acrobat (PDF) format. 128Mb 144/128 Megabit RDRAM Components To view our Rambus warranty, Click Here For orders of 100 or more, e-mail us at: rambus@kingston.com jd