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 : Formerly About Advanced Micro Devices -- Ignore unavailable to you. Want to Upgrade?

To: Tenchusatsu who wrote (67767)	8/5/1999 11:02:00 PM
From: Dan3	Read Replies (2) \| Respond to of 1575396

Re: bandwidth will still be less than that of Rambus DDR can be interleaved to gain bandwidth, but I doubt it's an issue for 98% of the computer market. It's latency where rambus lags, and it's latency that seems to have killed off the 300/600 rambus for computers (it's still a fine technology for video games). I don't have a specific link, but I recall reading several times that the 300/600MHZ rambus is, at best, no faster than standard (cas 3) PC100 in real world applications. PC100 has less than half the bandwidth of DDR 133, yet still provides similar performance to 300/600 rambus. Rambus 400/800 ups the bandwidth by 33%, instead of the 260% increase gained by moving to DDR. If bandwidth were the issue... But it doesn't seem to be. I've seen PC166 for sale at some of the performance sites - supposedly cas 2, but even at cas 3 latency is down to 16.5ns - as opposed to 30ns on PC100. PC133 at cas 2 gives 15ns latency, and that is what rambus must overcome. Meanwhile, rambus has had to binsplit the 400/800 into 40,45,and 50 ns parts - so it doesn't seem that we're going to see latencies less than 40ns soon (just as PCXX is cas 2 or cas 3, and cas 2 is the near term limit). The rambus memory bus makes up some of that, but if the 300/600 vs PC100 example is correct, it doesn't make up enough. The rambus datasheets are a little scary if you think of 50ns in the context of a 1GHZ part needing to go to main memory every 100 ns/clocks. Flip side is that none of this matters if AMD can't ship fast enough processors to pressure intel into providing full performance memory. If intel can stay 200MHZ ahead of AMD, rambus will do just fine, and if intel ships nothing but rambus chipsets, it's going to be a rambus future. It's easy to see why intel felt they had sue via to try to keep PC133 off the market. If via ships a chipset that supports the same processors as intel's chipset, it becomes possible to directly compare the two memory technologies and it's going to be hard to get people to pay a premium for a poorer perfoming product.