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: Bilow who wrote (28844)	9/7/1999 12:06:00 AM
From: Alan Bell	Read Replies (2) \| Respond to of 93625

Carl, I have read the documents on the Rambus site regarding Motherboard design issues and came to exactly the opposite conclusion that you have. And I have directed the design of a number of high speed boards using tools like Cadence Allegro and Spectra. Rambus recommends against critical signals making right-angle turns, but having many 45 degree turns will be unavoidable. The length of each 45 degree turn should be a minimum of 2mm to avoid a local impedance depression. The avoidance of 90 degree angles has been a standard part of high-speed PCB board design for over 15 years. I first saw it in the Motorola ECL Design guide in the early 80's. The difference in trace width as it makes a 90 degree angle creates an impedance discontinuity. But this isn't a problem because all high end design tools work with 45 degree angles. In fact, one has to go to extra effort to get a 90 degree angle. In many places, vias will drill through the ground and power planes, resulting in holes the planes. If these holes are close to each other and of sufficient size, they will overlap and effectively crate a slot in the plane. This is of particular concern near the small-pitch RDRAM footprints. Planes to which RSL signals are referenced must have no slots that impede the return current. Again, standard stuff. It is covered in the seminal book on PCB design - "High-Speed Digital Design : A Handbook of Black Magic". In fact, Rambus makes this easier by ordering the signals so they never have to cross over each other. They can go from controller to Rimm to Rimm without ever using a via. (If you create a slot with DDR it won't work either.) I could keep going but it would probably just fall on deaf ears. My conclusion is that designing a Rambus board, particularly with a complete design guide, isn't all that hard. -- Alan

To: Bilow who wrote (28844)	9/7/1999 12:56:00 AM
From: Bilow	Read Replies (2) \| Respond to of 93625

More on rambus alternatives... A lot of you know that IBM sold its discrete memory division recently. But they kept the embedded DRAM division. There is a reason for IBM having done this. The embedded technology is extremely important for the future. It completely avoids all the Rambus patents, and it is a lot faster. This is the technology that will eliminate the necessity for RDRAM at the small (in memory size) end. (There is already no need for rambus at the high end, as the AMD Athlon demonstrates, while there is a need for rambus at the low end, particularly dedicated game computers.) Here is a spec sheet on an IBM standard cell ASIC that includes embedded DRAM. Note the specifications on the DRAM: Random access: 13ns, 50MHz (kind of fast, isn't it?) Cycle time: 6.6ns, 160MHz (pretty slow, compared to RDRAM?) Data I/O: 256-bits (32 bytes, or 5GB/sec.) In short, just one of these embedded DRAMs is more than 3 times as fast as a rambus chip in random access, and provides over 3 times as much data bandwidth in burst mode. That's just one embedded DRAM. And it doesn't use any pins at all. And you can put more than one on a chip. Multiple macros per chip for greater capacity or functional flexibility. This stuff blows the doors off of rambus. It is new, it is going into designs right now, and it will be available in products you can buy within the next year, or possibly this year, and it is taking design wins from rambus right now. (pdf, page 381): chips.ibm.com I know that the people who believe in conspiracy theories are not going to believe the following Micron statement, but it does give you some sort of clue as to where DDR is being supported. Micron's statement on DDR SDRAM: When designing a memory subsystem requiring very high performance at a competitive price, plan to use DDR SDRAM for availability, pricing and overall market support. The 64Mb DDR SDRAM will offer a superior solution and a migration path from existing SDRAM designs. The DDR SDRAM is also supported by leading-edge controller manufacturers such as RCC and Opti, with more to follow. Memory manufacturers include Micron, Samsung, Hitachi, Hyundai, IBM, Fujitsu, Toshiba and Mitsubishi. Major systems companies like HP, IBM and SGI have announced plans to use DDR SDRAMs. micron.com Incidentally, Micron is estimating RDRAM as having a cost about 35% higher than DDR, with availability the same, and with total system bandwidth the same: Next-Generation DRAM Product Development micron.com -- Carl

To: Bilow who wrote (28844)	9/7/1999 10:37:00 AM
From: John Walliker	Respond to of 93625

Carl, I don't have time to respond to this post in detail right now, but I will make one point. Rambus have described how to ensure that their system can be made to work reliably. This is very good. Nothing about the PCB design rules is new, as you well know. Layout rules for high speed signals avoiding sharp corners were taught to electrical engineers well over 20 years ago. Maybe the box makers will have problems if they continue to churn out poor quality products. The good ones will adapt and be successful. Significant increases in computer speeds will not be achieved without careful attention to these matters. Higher performance does have a price, but it is not as high as you are suggesting. I know of plenty of PCB manufacturers who are happy to work to appropriate tolerances. John