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 (35232)	11/26/1999 3:53:00 AM
From: Tenchusatsu	Read Replies (2) \| Respond to of 93625

Carl, <As I posted before, IBM is working on an 8-way server in a MCM with 5000 pins.> And you think those 5000 pins are trivial to implement? Everyone at MPF who saw a sample of IBM's multi-chip module gasped because of its size and probable difficulty in manufacturing. Why do you think even the Alpha 21364 is integrating four RDRAM controllers onto a single processor core, instead of four DDR SDRAM controllers? Even Compaq/Digital is concerned about pin count with its not-so-high-volume Alpha. <Because of the above mentioned reduction in the cost of package pins, industry is no longer going in the direction of increasing bandwidth per pin. Instead, higher performance is being obtained through increased pin counts.> Oh, I think you'll be proved wrong in the near future. For example, AMD introduced a new LDT interface for enterprise servers which forms the interconnect for server platforms based on Athlon. Each LDT interface will consist of two unidirectional point-to-point buses, and each direction will be either 8, 16, or 32 bits wide. Each uni-directional pin runs at 1.6 Gbit/sec. Moreover, HotRail will also be implementing something similar in their server chipsets. Are they going down the wrong route? Furthermore, how come Intel's upcoming Timna is going to have an integrated RDRAM controller, instead of an SDRAM controller? Remember that Intel is aiming at a sub-$500 market with Timna. To be sure, the probable high cost of RDRAM at Timna's release could mean that Timna will initially support SDRAM via an MRH-S translator. But why is Intel leaving the RDRAM interface on Timna in the first place? <it now has only a 2 to 1 advantage over DDR-II. I know that Rambus longs say that DDR-II doesn't really exist, but this is silly. Nintendo got rid of Rambus by replacing the Rambus memory with a memory using a variation of the DDR-II specification.> First of all, we don't know whether Nintendo is using DDR, DDR-II, embedded DDR, or whatever. All we hear is scattered info. Second, we don't know when Nintendo's new Dolphin console will hit the market. Considering how late N64 was, I wouldn't hold my breath. And third, comparing Rambus to DDR-II is indeed silly. Rambus is here, now. I can call up Dell and order an RDRAM-based PC for under $2,000 if I wanted to. I can't even get DDR yet, much less DDR-II. <But don't take my view, look at the server business, which has removed the memory from its future almost universally.> You must have a very limited view then. Intel is still very committed toward pushing Rambus into all market segments, including servers. Though that push won't be happening as fast as Intel originally planned, everyone I've talked to at Intel said that DDR is still a shorter-term solution, and Rambus is the way to go. And no, I wasn't (just) talking to the guys who are paid to say that. Of course, you can just yawn and dismiss that last paragraph as "Yada yada, more Intel slogans from an Intel-a-bee." But it's too bad that I can't do much more on a public forum than repeat Intel's official line, because from my viewpoint, Rambus' future still looks rather bright to me, despite the attempts by the anti-Rambus coalition to rain on the parade. Tenchusatsu