It is always good to be back home after a whole week of interesting seminars, keynotes and meetings, as although our hotel
had nice bedding and great room service there’s no place like home. During my 9-hour flight back home I had quite some time to go over all of the datasheets and press information given to me during IDF and afterwards and here’s some other interesting information that didn’t make it to the previous columns.
I’m sure you remember the 'Rambus PC-1066 and PC-1200, pushing the envelope’
article I wrote a few months ago, in November of 2001 to be exact. That article was basically meant to investigate the performance and scalability of PC-1066 and PC-1200. I’m sure you’ve noticed that others have followed in my footsteps as other websites subsequently started to investigate this particular topic too. I think it wouldn’t surprise you if I told you Rambus had a working PC-1200 and PC-1066 system on display during IDF, of a similar configuration as the one we used during our testing.
Fig 1. The demo-system Rambus used during IDF, featuring a new Asus motherboard with support for 32-bit PC-1066 RIMMs.
In the article I briefly touched upon the new 32-bit Rambus modules that were under development at that time, and it may come as no surprise to you either that these are now production ready and Rambus actually had a demo-system featuring an Asus motherboard with the new 32-bit RIMMs up and running. The new RIMMs have a lot going for them as they significantly simplify the motherboard design by requiring less board space and fewer RIMM sockets. The new PC4200 RIMMs offer a high performance 4.2GB/s dual-channel memory solution in a single package that eliminates the need for two separate RIMMs as used in previous designs.
Fig 2. The new 32-bit RIMMs make efficient used of the unused space between the notches on a 16-bit RIMM as can clearly be seen from this picture.
But wait, wouldn’t a new RIMM module make things more complicated? New form factor, new sockets etc? Actually no, Rambus designed the new modules in such a way that they actually make efficient use of the unused space between the notches on a 16-bit RIMM, there’s enough space left to put in the second RDRAM channel. The result is a new RIMM module that is identical in size to the older RIMM module but does away with the need for a second RIMM in, for example, a Pentium 4 system.
The performance of the new module is of course no different from the old approach that used two 16-bit modules, from a cost and space-constraints point of view however this is a significant improvement. I’m sure many manufacturers of small footprint PCs, for example using the Micro-ATX formfactor, will welcome the new 32-bit RIMMs as it saves them board space and thus allows them to build more functionality in their PCs, or simply make them smaller and more cost efficient overall. The new 32-bit module might well mark the debut of RDRAM in small desktop PCs that before wouldn’t be able to accommodate four RIMM-sockets due to space constraints.