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: jim kelley who wrote (50995)	8/24/2000 2:15:53 PM
From: Bilow	Read Replies (2) \| Respond to of 93625

Hi jim kelley; Re Newell's credentials. It really doesn't matter, but I would like to note that you still haven't provided a link. I provided you with the search engines required, since you still come up zero, I assume that you have to admit that the guy doesn't work for Johns Hopkins. I've forgotten what the silly argument the guy provided was. It is quite pointless to dispose of it, as the vast majority of the people who read this thread are in no position to understand the arguments involved. But bring it back, maybe I'll take a look at it. (This is an invitation for you to repost that useless garbage on the thread. Go for it!) Re soldering memory on the motherboard, and why it hasn't happened before... One of the long term trends in the industry is a reduction in the average number of chips carried in the typical (i.e. mainstream) PC. This fact gets the Rambus longs hot because of the importance of "granularity" in memory, so this should be a familiar trend. But granularity in DIMMs (and RIMMs) is limited by the cost of the DIMM (RIMM) PCB. For this reason, despite the fact that you could manufacture an RDRAM RIMM with as little as a single memory chip on it, you don't see RDRAM RIMMs with fewer than 4 chips soldered onto it. That is also (more or less) the limit for DDR/SDRAM DIMMs. Note that using the x32 parts, DDR DIMMs could be produced with as few as two memory chips per DIMM, but I doubt that it will ever happen. Instead, when it reaches the point that the typical DIMM/RIMM would end up with only 4 or less memory chips, the industry will magically and suddenly switch over to the expedient of soldering them directly on the motherboard. Like most things in engineering, it's all about money. Putting chips into DIMMs costs money, but you get tangible benefits in return: (1) You don't have to add the memory until it is time to ship the motherboard (or you can even add them in the field). This means less money tied up in the board and better JIT use of facilities and capital. (2) The motherboard can later be configured with more or less memory, faster or slower, according to the customer desires. (3) The customer can add memory later, as an upgrade to his system. (I've noticed that I do this less and less in recent years. Instead, I just get the system with enough memory to begin with.) (4) I suppose I could add that you gain more room on the motherboard. All these advantages will still exist when the industry switches over to soldering memory on the motherboard, but like everything else in engineering (well almost everything) it is a matter of economics. As the granularity problem gets worse and worse it becomes more and more pointless to allow the user to specify his memory. (The user ends up with fewer and fewer choices, so why bother?) Since the cost of memory chips is more or less constant, per chip, as memory systems use fewer and fewer chips the total dollars going into the memory system gets less and less. In that situation, the cost of those IMMs gets to be a larger and larger percentage of the cost of the machine. Also, as frequencies continue to increase, the amount of maximum performance that is lost by going to a IMM will increase. All these things suggest that eventually memory will be soldered directly on the motherboard. A link from Samsung showing this trend: The platforms will use a few 128-Mbit SDRAM chips or even fewer 256-Mbit Direct Rambus DRAM devices, he added. Costs will be reduced further because no memory modules will be required. ebnews.com But that's not all. Eventually, memory will come packaged with the processor, in the same SOC. -- Carl