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.

Politics : Ask Michael Burke -- Ignore unavailable to you. Want to Upgrade?

To: Michael Bakunin who wrote (77802)	3/14/2000 9:28:00 AM
From: gnuman	Read Replies (1) \| Respond to of 132070

mb, re: <..but, isn't one of Rambus's selling points a similar interleave of multiple channels?> I believe it's true with the 840 chip set you can interleave two channels. (Which can result in double the data rate). I understand without interleave RDRAM 800 only provides double the data rate of PC100 DIMM. (RDRAM R/W's 2 bytes vs DIMM R/W's of 8 bytes). Also, DDR equals the data rate of RDRAM and could be a server solution. But they have timing issues similar to RDRAM. I think the real issue is cost. By interleaving multiple DIMM channels you can also increase the data rate and I think the price/performance of that solution is more attractive to the server industry. (Interleaving two DIMM channels equals the data rate of non-interleaved RDRAM). And for the desk top RDRAM is currently much too expensive. As the cost of 256Mb RDRAM's come down this could change. (You can make a 32MB module with one 256Mb RDRAM). But the mobo/RIMM designs have become more science than art and require higher mfg/test costs than DIMM's. (I still think there will be problems with mixing/matching RIMM's from different suppliers). If you haven't seen it, Toshiba has a white paper on the merits of the various DRAM technologies. (Biased to RDRAM). toshiba.com JMHO's

To: Michael Bakunin who wrote (77802)	3/14/2000 8:50:00 PM
From: Scott D.	Respond to of 132070

re: Interleaving Rambus Yes, Rambus is 1/4 the width of DIMMs, which makes it possible to have interleaving with a smaller minimum configuration. The Intel 840 has two Rambus channels, I believe, and allows interleaving on workstations. Rambus has an advantage of making interleaving possible on a smaller minimum memory configuration than DIMMs. So paying a premium might make since on a workstation design. But Rambus is a long way off in the server world, I think, because of the additional module cost. Take an Itanium server maxed out with 64Gb and figure the Rambus cost premium. It is going to increase the cost substantially with no apparent benefit to the customer (they get interleaving either way). I do not really follow the Rambus story too much and cannot remember why anyone would prefer it. Except Intel, who saves money by reducing pin count.