Intel admits SDRAM as good as Rambus memory

A product change notification (PCN) from Intel has supplied more details about the difficulties with three motherboards which we exclusively revealed a while back. And other documents we have seen show that Intel believes that synchronous memory can perform as well as Rambus. PCN 943 is a document Intel issued to its customers ten days ago which includes information on the errata which is causing it to withdraw two 840 mobos and one 820 mobo. The document says that after making a stepping to the chipset, the corrected technology will be made available on May 17th next. It will fix erratum number one in the MRH-S (memory repeater hub), and will issue a corrected MRH-S design to enable (fix) dual SDRAM repeaters for each channel. Stepping B0 will be pin to pin compatible with Intel's i840 chipset MRH-S A0, and so there is no need of a board "re-spin", in the document's quaint words. But there will be the necessity for a BIOS and minor driver revision, which means that there could well be a requirement for WHQL re-certification. In news from another Intel document, we learn that SDRAM is a good thing. A white paper Intel issued on the 840 says that the 82804 SDRAM based MRH can give bandwidth equal to Rambus. The document says: "In server applications, memory capacity is typically even more important than memory performance. Large memory subsystems can be divided into many banks, allowing the main memory to be interleaved. This approach delivers higher performance in multithreaded and multiprocessing environments. These large memory systems can offer bandwidth equal to that of RDRAM although implemented with SDRAM." (Our italics). In yet another Intel document we have seen, it is apparent that the company realised it could go with PC-133 and DDR (double data rate) memory as far back as October/November last year. This document, called the PC SDRAM Specification, makes the following statement: "The objective...is to define a new Synchronous DRAM specification ("PC SDRAM") which will remove extra functionality from the current JEDEC standard SDRAM specification, so that it will be a fully compatible device among all vendor designed parts. It should be easy to design and manufacture and highly cost optimized for the main stream volume desktop Intel architecture PCs". So where does that leave Rambus memory? A reader has forwarded to us a piece from The Street, by Fayad Abbasi, and dated last Friday, which says: "Short-term market dynamics have taken over the trading of Rambus (Nasdaq: RMBS). We recommend investors sell shares at current levels and wait for normal market conditions to return." Rambus' share price closed at $204 on Nasdaq last Friday, and we received the following nice email from a reader. "Just a quick note to say thank you for tipping Rambus back in December at $69. I just sold them all for 200-and something percent profit - no doubt I'll be kicking myself in a couple of weeks, what with the Rambus memory conference and PS2 launch next week and as yet unannounced communications products, but at least the taxman won't quite be getting his thieving mitts on my dosh this tax year." The Register would like to re-iterate the fact that we do not invest money in companies we write about. ® See also Three Intel mobos scrapped because of chipset problems More Intel codenames to learn Intel Developer Forum Spring 2000