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: Skeeter Bug who wrote (81410)	3/6/2002 4:24:23 PM
From: JD_Canuck	Read Replies (2) \| Respond to of 93625

Skeeter: I do take the caution to try to separate emotional need to be supported in my thinking with being open to changing environments...with thanks. What i see is a huge level of anger against RMBS from within the industry...and i still can't see the basis for it. From where i sit...it looked more like resentment to change than reaction to blackmail. If i have my info right...those European patent applications predated Jedec, they were supported by the only patent experts to examine the claims to this point, as applying to not only DDR but also SDRAM. RMBS had SOMETHING that caused them to be invited to sit in on Jedec. Now, the industry(or some people in it) claim RMBS had nothing...that all the claims were stolen from their Jedec participation. This just does not fit the facts that i have put together. If i am wrong on these premises...please point out where. Heck...i am searching for answers here, not pushing my decisions on anyone. As to INTC being upset...at what? I know of no area where RMBS did not pull off what they agreed to contribute. Now INTC being upset at MU...that i can understand. i know i would never turn my back on a company that entered into an agreement, accepted the cash up front, and then caused a serious delay in my business plans as they refused to provide what they had agreed to...forcing me to go to another source much later, and lose market share as a result. INTC lost market share due to the lead that AMD was able to put together during this hi cost RDRAM period, not because RDRAM was not capable.

To: Skeeter Bug who wrote (81410)	3/6/2002 6:11:43 PM
From: Bilow	Read Replies (2) \| Respond to of 93625

Hi Skeeter Bug; Re: "the only question i have is whether rdram offers a significant enough cost/benefit ratio at the much higher mhz levels." With 2GHz processors, a single channel of PC2100 gets a processor within a couple percent of performance equal to two channels of PC800. As processors continue to speed up, they will simultaneously continue to increase their on chip cache memory sizes. To some extent, this will decrease the need for higher speed interfaces. But let's ignore this effect completely and look at what bandwidth could be provided with current DDR technology. First, as the Rambus people have repeatedly noted, going into the future granularity becomes more of an issue. But this is not something that is entirely in favor of Rambus, especially given the fact that DDR already provides much higher bandwidth per chip than RDRAM. One of the other effects of granularity is that it makes it more and more feasible to design computers that use point 2 point memory interfaces, and have no user serviceable memory. This is not much of a stretch, given that the graphics industry (which tends to lead the PC industry in memory technologies) went through the same changes years ago. Originally, graphics memory was user installable. Then they switched to soldered in memory, and if you want a board with more memory, you have to make that decision at the time you purchase the board. This allowed much higher frequencies. So what kind of bandwidth / memory size can we imagine that they will attach a mainstream processor in the future? Already processors are being prototyped with 128-bit wide integrated DDR interfaces (i.e. SledgeHammer). Already Samsung is sampling 800Mbps DDR chips. If you combine these, you get a bandwidth of .8 x 128 / 8 = 12.8GB/sec. That is 6 times as fast as a single PC2100 channel, so we should be able to get to processors at around 12GHz or so. But advances in packaging will probably increase the memory width from 128 bits to 256 bits, and over the same time period they'll go from DDR to DDR2. With DDR2 giving at least 1200 Mbps and with 256-bit wide memory interfaces, we get memory bandwidths of 38.4GB/sec, which should handle processors up to around 40GHz. After that time, the next step will be to bring the memory onto the same "multi chip module" (or something similar). That will allow the memory width to immediately increase by another factor of four (or more) to 1024 bits, and will increase the data rate from 1.2Gbps to at least 2Gbps, and probably as high as 10Gbps. With 1024 bits and 2 to 10Gbps, our memory bandwidth increases to 256 to 1280GB/sec, which will be enough for processor speeds up to 250 to 1000 GHz. At that time separate memory chips may be removed entirely from desktop processors, and instead included with the processor chip itself. That's what happened in the pocket calculator market, and it also happened in the low end graphics card market. With all the connections on chip, there will be no need to use Rambus at all. -- Carl