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.

Non-Tech : The Critical Investing Workshop -- Ignore unavailable to you. Want to Upgrade?

To: Sully- who wrote (30989)	8/29/2000 7:23:01 PM
From: Sully-	Read Replies (1) \| Respond to of 35685

Driving the RAM-bus The Bear Argument By Rob Landley (TMF Oak) The computer industry is driven by Moore's Law, which states that the bang for the buck you get buying new computer hardware doubles every 18 months. Rambus is a company that makes fast memory for computers, and its problem is that Moore's Law has now caught up with it. Rambus didn't beat Moore's Law. It made faster RAM chips than its competitors at the time, but it was considerably more expensive as well. Intel used to be the biggest backer of Rambus memory, because Intel's new 64-bit "iTanium" chip design is a real memory-speed hog. While Intel's older 32-bit Pentium class designs used instructions ranging from 8 bits to 32 bits, Intel's newest designs use fixed-length 64-bit instructions. This means a Pentium could sometimes read in two or three smaller instructions instead of one big one, but iTanium needed to read in twice the memory to get each instruction, and never could read in multiple smaller instructions in one gulp. This meant that under the right conditions, a Pentium could literally outperform Intel's newer designs because reading instructions in from memory was such a big bottleneck. (This is why iTanium's ship date keeps slipping: its performance stinks.) To compensate, Intel wanted to throw money at the problem and use ultra-fast Rambus RAM in iTanium systems. (Of course Pentiums could benefit from faster RAM too, but this detail could be safely overlooked by Intel's marketing department.) That was a year ago. Times have changed. Intel has had chronic problems making stable and affordable interface chips to allow motherboards to use Rambus memory. But far worse for Rambus, the commodity SDRAM other manufacturers are producing has caught up with the speed of Rambus memory. Intel's own test results show that commodity PC 133 SDRAM slightly outperforms Rambus memory, and is way cheaper. As a result, leaked internal Intel documents indicate Intel is drastically reducing its commitment to Rambus memory in favor of the cheaper commodity SDRAM from other manufacturers. As early as October, Japan's largest chipmaker, NEC, halted production of Rambus chips and turned that manufacturing capacity over to producing SDRAM. Rambus has responded to all this by getting patents on blindingly obvious things that a sane patent office wouldn't grant a patent on, and then suing SDRAM manufacturers for infringing on these patents. The worst of the lot is probably the "dual edge triggering patent," which even non-technical people can see is not something they should have gotten a patent on. It works like this: Computer circuitry runs off of a clock, meaning a series of timed electrical pulses generated by a vibrating crystal. The signal goes on and off and on and off, like somebody flipping a lightswitch, letting the circuit know when everything it's done should be finished and it's OK to go on to the next step. If part of the circuit finishes early and the clock signal is still on from last time, it might do two things in one clock cycle and move on to the next thing before the rest of the circuit is ready. (This is bad.) So instead of being "level-triggered," which means the circuit notices that the signal is on, most clock-driven circuits are "edge-triggered," which means the circuit's cue to go is when the signal changes. The "leading edge" is where it switches from off to on, and the "trailing edge" is where it switches from on to off. Picking one of those edges as a trigger for the circuit is a better design than just looking if it's on or off, because the signal could stay on for a while. But when you use edges, there's no room for confusion about when stuff should happen. Rambus does not have a patent on any of this. It's obvious stuff any electrical engineer learns as an undergraduate. The rate at which signals can be sent down wires depends on the size, shape, length, materials (and so on) of the wire in question. Faster signals mean shorter pulses, and beyond a certain speed the ultra-short pulses are too fuzzy to be legible (due to Doppler effects, interference from neighboring wires, and just plain shortness). For a long time this wasn't a problem, but these days your clock can only go just so fast before you have to rewire your circuit and try to make it smaller and clearer. Of course there's an easy way to double your circuit speed. Trigger the circuit on BOTH edges. (Trigger when the clock signal switches on, trigger again when it switches off.) Back when you could just plug in a faster clock if you needed one, this wasn't worth the trouble, but now it's an obvious little performance trick. Except Rambus patented it, and is suing everybody else using that trick. And they're suing the CUSTOMERS of everybody who uses that trick. Needless to say, the engineering community at large is extremely annoyed. They didn't steal that idea from Rambus. It's darn obvious to the people who do this sort of thing for a living. This is why the DRAM industry as a whole is considering filing a class-action antitrust suit against Rambus, for abuse of ridiculous patents and general bullying tactics by a sore loser. This is not a company I would put money into. aolsnapshot.fool.com