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: DownSouth who wrote (17424)	3/17/1999 7:24:00 PM
From: Dave B	Read Replies (1) \| Respond to of 93625

Found this while doing some research today. It sounds bad at first, but keep in mind that ANY memory technology that would satisfy the need for faster memory (to keep up with the processors) would run into these issues. _________________________ VLSI Research March 7, 1999: From the front lines . . . The Rambus conference, held last week in Japan, was a hotbed of controversy. The problems are pretty basic: They can't build it, can't package it, and can't test it -- at least cost effectively. The public consensus is that the MicroBGA package is dead for Rambus. Rambus, the company, has backed off of their endorsement of the MicroBGA package and is cautious about endorsing any new package. The delays are a real plus for the late-to-market innovators who have better ideas. The new FormFactor micro-spring package is generating a lot of excitement. FormFactor showed they had developed a dual-sided RIMM version. It could be a real winner from a cost and speed vantagepoint. However, they must still have to win over those hard crusted manufacturing guys in the 'I heard that before' camp. Toshiba showed data indicating that Advantest's T5591 could achieve +or- 40PS (picoseconds) accuracy correlation across eight test sites. But it is questionable as to when anyone will be able to build a tester in volume that is accurate enough to test Rambus. If you can't test it, you don't know if you can build it. Edge Placement Accuracy (EPA) will be the battleground for Rambus testing. If you think about building the industry around a memory at 800MHz, when its best production microprocessors run at 450MHz, you quickly come to the conclusion that it is a pretty tall order. 800MHz means that each clock cycle will happen every 1250 picoseconds (PS). To put this in more comprehensible terms, if one second could be expanded into an eighty-year lifetime, it would mean birthdays would come every three seconds. Normally the tester manufacturer is asked to build a system that can measure the edges of each cycle, or EPA, to within 7% of half a clock cycle. With an 800MHz clock, the tester must be able to measure these cycles within 43PS. SO WHAT'S THE BIG DEAL? It's about yield, which means its about money. If you do the math, you'll find that yield should be roughly a linear function of Edge Placement Accuracy. The reason is that guard-banding of the tester will cause it to bin good parts as bad if the tester can't measure it. So, your typical 80PS EPA accuracy Rambus tester will bin 5.8% of the parts it tests as bad, even though they are good. Since the tester can test about 10M units a year, it means that 580 thousand, perfectly good, parts will be tossed each year. So, a new tester, costing $2.5M, will pay for itself in just over a quarter. By the end of the year, using a $14 chip price, it bring an additional $5.6M to the bottom line. In other words, it adds $561 of pure, unadulterated profit to each 64Mb wafer processed. Is it real? One manufacturer did a comparison between testers with 80PS and 50PS EPA and found gained 6.5% in yield. Fortunes will be made in this market. _____________ Dave B