Browse Prior Art Database

SRAM Cube Application for High Performance Workstations

IP.com Disclosure Number: IPCOM000111942D
Original Publication Date: 1994-Apr-01
Included in the Prior Art Database: 2005-Mar-26
Document File: 2 page(s) / 36K

Publishing Venue

IBM

Related People

Daniels, L: AUTHOR [+3]

Abstract

High performance workstations require high capacity cache memories requiring a large number of chips. Wireability constraints that result from such a memory arrangement bring about long multi-drop nets (typically 4 and 8 drop), which in turn degrade signal integrity resulting in negative performance impact.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 90% of the total text.

SRAM Cube Application for High Performance Workstations

      High performance workstations require high capacity cache
memories requiring a large number of chips.  Wireability constraints
that result from such a memory arrangement bring about long
multi-drop nets (typically 4 and 8 drop), which in turn degrade
signal integrity resulting in negative performance impact.

      To eliminate these problems, this invention suggests combining
the cache chips into a "memory cube".  Use of SRAM in this format
reduces the board requiring by more than an order of magnitude.
Additionally, because the address line wiring is provided on the
bottom surface of the cube, address line length between the chips is
reduced to (essentially) zero.  The stubs between the chips are also
eliminated.  Thus, a long multi-drop net becomes a short
point-to-point (although still multi-load) net, which has a far
better signal integrity and shorter net delay.  Further, the desired
result is achieved utilizing a far more trivial line termination
scheme.

      However, the power density (not power dissipation) for the cube
is much higher than for the case of individual chips mounted on a
carrier.  This is a problem that must be addressed during design of
the processor.  Some of the additional area that is freed up by use
of the cube technology can be used to reduce he thermal problem
(larger heat sink, etc.).  It is not expected, however, that all the
area will be required to attain th...