To: Darrell D. Conrade who wrote (7057 ) 3/1/1998 4:19:00 PM From: Rob S. Respond to
PC hardware has gotten more powerful over the years because of advances in uP power - cramming more and more "transistor equivalent circuits" on a given size piece of silicon. The motherboards (MB) that these and the support chips go into have tried to keep up with the uP chips by increasing the frequency (speed) and bandwidth (number of lanes of traffic) of information that travels between each component or subsystem, such as the uP and the graphics & sound cards and disk controllers. The PCI bus used in most systems today increased the bandwidth but was limited in frequency. The sockets that are used to plug processors into the motherboards have also increased in number of pins (bandwidth) and frequency capability. The socket 7 MB architecure and PII PCI bus has been extended to 60 Mhz for Pentium chips and to 75 Mhz for boards that support Cyrix processors. AMD, Cyrix and IDTI are supporting "Super Socket 7" which will extend the MB frequency further to 100 Mhz. Increasing to this frequency is a difficult task. All the components used in the system that fit on the bus are being stressed to very tight timing requirements. At the higher frequencies, short circuit traces on the motherboard look increasingly distant in terms of controlling the timing of when signals get to where they need to go. What's called "reactances", the inductance and capacitance of the traces and component interconnects, look increasingly like detours or roadblocks to the travel of information. Some of these roadblocks are due to the nature of the materials used in circuit boards - the glass epoxy MB and copper traces are good to fabricate complex circuits but are less than ideal from the standpoint of limiting the reactances that create traffic problems. One way to reduce the problems is to pack components closer together. But thats difficult to do using traditional MBs. Intel saw that at some point the problems of trying to increase MB frequency would become counter-productive in terms of manufcturing cost and complexity. One solution to the overal roadblock in ramping speed is to put chip components onto materials that have better transmission properties (lower reactances and finer trace lines) and can be packed closer together than conventional MB materials. An approach that has been used extensively starting with military/aerospace applications is called "multichip modules". Of course for those applications cost was hardly ever the defining requirement. So when Intel was confronted with the entry of AMD and Cyrix as worthy competitors, they shifted the MB design toward the new slot 1 and soon slot 2 designs that use a commericalized version of the multi-chip modules approach. Slot 1 or similar designs that may evolve, mount the uP and SRAM onto a substrate material that has better conductive properties and on which the parts can be packed closer together. This reduces the bottlenecks between the components on the slot card. You still face similar limitations when you go from the slot connector on the MB to other components.
