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: Dan3 who wrote (29543)	9/15/1999 4:15:00 AM
From: Tenchusatsu	Read Replies (2) \| Respond to of 93625

Dan, <Rambus has an advantage only if there is a requirement for streaming memory.> Dan, you're forgetting about AGP. Cards like the popular Riva TNT2 actually take advantage of the bandwidth that AGP-2x provides. And that bandwidth can be much more than "streaming data from disk or network feeds." It all boils down to latency and bandwidth, as you know already. AGP and PCI is more dependent on bandwidth than latency, since they tend to transfer large bursts of data at one time in order to fill their buffers. The processor is more dependent on latency, since shorter latencies means the CPU stalls less on cache misses. If your memory gets bogged down by the large burst transfers coming from AGP and PCI, the CPU latency increases by a sizable amount. (Just ask Dell.) RDRAM does a much better job handling this kind of traffic. That's why I keep saying that RDRAM is better able to handle concurrent traffic from the CPU, AGP, and PCI. The effects might be noticeable only in a few situations at present, but if history is a reliable guide, it will become much more noticeable in the future. <PCXXX can burst a cache line as quickly as Rambus, VC DDR more quickly, and the non rambus memory types can always be interleaved if someday, some application comes along that really does need that kind of streaming memory performance.> First, PCXXX SDRAM does not burst a cacheline as quickly as Rambus. Sorry, that's just a completely ignorant statement. Second, I never even heard of the combination of VC SDRAM and DDR SDRAM. Do you know how virtual channel technology works? Why hasn't the anti-Rambus coalition rallied around VC? Why can't the chipset duplicate the benefits of VC technology using proper buffering and prefetching? And third, do you know how expensive it is to interleave any type of memory, including SDRAM? The 460GX, a 4-way Merced chipset, interleaves SDRAM, but that's OK since this chipset is not meant to be cheap. Plus, 460GX has to support huge amounts of memory anyway, which is another benefit of interleaving. But for systems that won't cost you the equivalent of a Lexus, interleaving is not a very cost-effective solution. Tenchusatsu