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: Bilow who wrote (32968)	10/26/1999 9:35:00 PM
From: richard surckla	Respond to of 93625

TunnelMos(TM) Memory? Now what's new?.... national-scientific.com national-scientific.com

To: Bilow who wrote (32968)	10/26/1999 9:38:00 PM
From: wily	Read Replies (1) \| Respond to of 93625

>>P.S. Rambus also provides much, much less performance per NC pin.<< Thankyou for the analysis Mr. Not Connected.

To: Bilow who wrote (32968)	10/26/1999 9:50:00 PM
From: Dan3	Respond to of 93625

Re: Rambus also provides much, much less performance per NC pin. :) <ROTFLOL> Now they know where to focus development attention! Dan

To: Bilow who wrote (32968)	10/26/1999 10:50:00 PM
From: Tenchusatsu	Read Replies (2) \| Respond to of 93625

Carl, the number of signal pins on a Rambus channel is actually one half of what you stated. The document you posted from www.rambus.com is actually a RIMM spec sheet. And as you know, a RIMM has signal pins going in and signal pins coming out. Therefore, there are actually 37 signal pins on a Rambus channel, not 74 as you might think. So to redo your calculations: The bandwidth per signal pin on the DDR module is 2.128GB/sec / 125 = 17MB/sec/pin. The bandwidth per signal pin on the Rambus module is 1.6GB/sec / 37 = 43MB/sec/pin. Seems like Rambus is pushing a lot more bandwidth per signal pin than DDR SDRAM. This doesn't translate into pin savings on the RIMM because of the two-ended channel, but it does translate into pin savings on the memory controller. Tenchusatsu

To: Bilow who wrote (32968)	10/26/1999 11:28:00 PM
From: TigerPaw	Read Replies (2) \| Respond to of 93625

The Rambus channel pins both enter and exit the RIMM The key is not saving pins on the RIMM module, it's about saving pins on the memory controller. TP

To: Bilow who wrote (32968)	10/27/1999 6:49:00 AM
From: Bilow	Read Replies (1) \| Respond to of 93625

Hi all; Further notes on RDRAM vs DDR SDRAM. Pin Costs. Back on September 29, on this thread (see #reply-11393109 ), I put together a set of 10 links with quotes on the subject of pin costs. There were quite a few estimates for what a pin costs, and they ran all the way from 0.2 to 3.0 cents per pin, depending on the technology, and whether the estimate included the cost of assembly as well as just the package costs. Rambus used the 3 cent per pin technique, all the other techniques maxed out at 2 cents per pin. Given this fact, a good bit of the savings advantage of Rambus disappears in the higher cost of its package, but I expect that this higher cost will eventually come down. I see no reason why RDRAM I/O should cost more to "bond" then other techniques, unless it is due to the very high current requirements of RDRAM (i.e 90mA). Because of my uncertainty, I am going to assume that Rambus pads cost the same as other pads. In fact, it is certain that they are no cheaper, and likely that there are higher costs associated with controlling the impedance &c. Now, given the number of signals (see #reply-11714426 ) required on the two technologies, we can compute the total system costs for the two choices, DDR SDRAM vs RDRAM. The cost to the controller chip for a Rambus path is 37 x $0.02 = $0.74 per channel for Rambus. For DDR SDRAM, the corresponding figure is 125 x $0.02 = $2.50 per channel. The difference in cost $2.50 - $0.74 = $1.76, seems like a pretty high amount, but I have made much worst case assumptions all around. In the future, as packaging technologies improve and the costs per pin go down, this number will go down as well. Now we need to compare the cost difference per DIMM vs. RIMM module. We already know that the RIMM compatible motherboards need to be scrapped, uhhh, I mean more carefully designed and manufactured, but there are other costs that are easier to quantify. For instance, the heat sink, uhhhh, "spreader", costs $1.56 per RIMM: The price of $1.38 each, in quantities of 25,000, includes two Cool Covers with interface material. Brass Push Pins are an additional 18 cents per set. techweb.com At this point, we are down to a cost difference of 20 cents per memory channel. This is going to be considerably below the cost of one of those RIMM spacers that they have to insert when there is no memory installed in a slot. The costs of all the other things are going to add up quickly. The Rambus royalty, for instance. Assembling the more complex RIMM module requires more labor, and also requires more expensive testing, as has been extensively mentioned in the industry press. The upshot to all this is simple: RDRAM just isn't going to save anyone any money. It is not a cost effective technology now, nor will it ever be. The cost differences with DDR SDRAM will get worse in the future, not better, as the cost of all that hardware stays constant while the Moore's Law equivalency for packaging drops the cost of SDRAM pins. -- Carl P.S. Sorry, Wily, I have to be careful on this thread, some of these people don't appreciate my sense of humor.