International Business Machines Corporation

David J Perlman

Claude Louis Bertin

John Edward Cronin

Wayne John Howell

Kenneth E Beilstein, Jr

Kevin Charles Gower

Mark William Kellogg

Christopher P Miller

Bruce G Hazelzet

James Marc Leas

Edmund Juris Sprogis

Gustav Schrottke

Gordon Arthur Kelley

Jody John VanHorn

Dennis Charles Dubois

Paul A Farrar

Robert J Von Gutfeld

William Paul Hovis

Praveen Chaudhari

Akella V Satya

Robert M Blake

Howard L Kelly

Helen Janet Kelly

Eugene E Tynan

Hans Kolan

John B Kiessling

Kenneth M Hansen

A high reliability dual inline memory module with a fault tolerant address and command bus for use in a server. The memory module is a card approximately 151.35 mm or 5.97 inches long provided with about a plurality of contacts of which some are redundant, a plurality of DRAMs, a phase lock loop, a 2 or 32K bit serial EE PROM and a 28 bit and a 1 to 2 register having error correction code (ECC), parity checking, a multi-byte fault reporting circuitry for reading via an independent bus, and real time error lines for determining and reporting both correctable errors and uncorrectable error conditions coupled to the server's memory interface chip and memory controller or processor such that the memory controller sends address and command information to the register via address/command lines together with check bits for error correction purposes to the ECC/ Parity register. By providing the module with a fault tolerant address and command bus fault-tolerance and self-healing aspects necessary for autonomic computing systems compatible with industry-standards is realized. The memory module corrects single bit errors on the command or address bus and permits continuous memory operation independent of the existence of these errors and can determine any double bit error condition. The redundant contacts on the module prevents what would otherwise be single points of failure.

High reliability memory module with a fault tolerant address and command bus

A high reliability dual inline memory module with a fault tolerant address and command bus for use in a server. The memory module is a card approximately 151.35 mm or 5.97 inches long provided with about a plurality of contacts of which some are redundant, a plurality of DRAMs, a phase lock loop, a 2 or 32 K bit serial EE PROM and a 28 bit and a 1 to 2 register having error correction code (ECC), parity checking, a multi-byte fault reporting circuitry for reading via an independent bus, and real time error lines for determining and reporting both correctable errors and uncorrectable error conditions coupled to the server's memory interface chip and memory controller or processor such that the memory controller sends address and command information to the register via address/command lines together with check bits for error correction purposes to the ECC/Parity register. By providing the module with a fault tolerant address and command bus fault-tolerance and self-healing aspects necessary for autonomic computing systems compatible with industry-standards is realized. The memory module corrects single bit errors on the command or address bus and permits continuous memory operation independent of the existence of these errors and can determine any double bit error condition. The redundant contacts on the module prevents what would otherwise be single points of failure.

A high reliability dual inline memory module with a fault tolerant address and command bus for use in a server. The memory module is a card provided with a plurality of contacts of which some are redundant, a plurality of DRAMs, a phase lock loop, a 2 or 32K bit serial EE PROM, and a 28 bit 1 to 2 register having error correction code (ECC), parity checking, a multi-byte fault reporting circuitry for reading via an independent bus, and real time error lines for determining and reporting both correctable errors and uncorrectable error conditions coupled to the server's memory interface chip and memory controller or processor such that the memory controller sends address and command information to the register via address/command lines together with check bits for error correction purposes to the ECC/Parity register.

A high reliability dual inline memory module with a fault tolerant address and command bus for use in a server. The memory module is a card approximately 151.35 mm or 5.97 inches long provided with about a plurality of contacts of which some are redundant, a plurality of DRAMs, a phase lock loop, a 2 or 32K bit serial EE PROM and a 28 bit, a 1 to 2 register having error correction code (ECC), parity checking, a multi-byte fault reporting circuitry for reading via an independent bus, and real time error lines for determining and reporting both correctable errors and uncorrectable error conditions coupled to the server's memory interface chip and memory controller or processor such that the memory controller sends address and command information to the register via address/command lines together with check bits for error correction purposes to the ECC/Parity register. By providing the module with a fault tolerant address and command bus fault-tolerance and self-healing aspects necessary for autonomic computing systems compatible with industry-standards is realized. The memory module corrects single bit errors on the command or address bus and permits continuous memory operation independent of the existence of these errors and can determine any double bit error condition. The redundant contacts on the module prevents what would otherwise be single points of failure.

Methods and apparatus are set forth for burn-in stressing and simultaneous testing of a plurality of semiconductor device chips laminated together in a stack configuration to define a multichip module. Testing is facilitated by connecting temporary interconnect wiring to an access surface of the multichip module. This temporary interconnect wiring electrically interconnects at least some semiconductor device chips within the module. Prior to burn-in stressing and testing, a separate electrical screening step occurs to identify any electrical defect in the connection between the temporary interconnect wiring and the multichip module. If an electrical defect is identified, various techniques for removing or isolating the defect are presented. Thereafter, burn-in stressing and simultaneous testing of the semiconductor chips within the multichip module occurs using the temporary interconnect wiring. Various alignment and test fixtures are described for facilitating this burn-in and simultaneous testing of the semiconductor chips within the multichip module.

Methods and apparatus for burn-in stressing and simultaneous testing of

A fabrication method and resultant monolithic electronic module having a separately formed thin-film layer attached to a side surface. The fabrication method includes providing an electronic module composed of stacked integrated circuit chips. A thin-film layer is separately formed on a temporary support which is used to attach the thin-film layer to the electronic module. The disclosed techniques may also be used for attaching an interposer, which may include active circuity, to an electronic module. Specific details of the fabrication method, resulting multichip packages, and various thin-film structures are set forth.

Method and workpiece for connecting a thin layer to a monolithic

Methods for alignment of stacked integrated circuit chips and the resultant three-dimensional semiconductor structures. A thickness control layer is deposited, as needed, on each integrated circuit chip. The thickness of the layer is determined by the thickness of the chip following a grind stage in the fabrication process. Complementary patterns are etched into the thickness control layer of each chip and into adjacent chips. Upon stacking the chips in a three dimensional structure, precise alignment is obtained for interconnect pads which are disposed on the edges of each integrated circuit chip. Dense bus and I/O networks can be thereby supported on a face of the resultant three-dimensional structure.

Process for controlling distance between integrated circuit chips in an

Process for aligning etch masks on an integrated circuit surface using

Cube wireability enhancement with chip-to-chip alignment and thickness

A cube package of stacked silicon semiconductor chips. To accommodate cube packaging, a metal transfer layer is added over the passivated chip face to bring all of the surface electrical contacts to a common chip edge. The metal transfer layer is insulated from the chip face and from the adjacent chip in the stack by polymer layers having a low dielectric constant, and a thermal expansion coefficient matching that of the stacked chips. An adhesive polymer layer is added to strengthen the bond between the first polymer layers and the adjacent chip in the stack, by deposition of the adhesive layer and partial cure at the wafer level, and then full cure when the chips are stacked together to form the cube.

Polyimide-insulated cube package of stacked semiconductor device chips

A method of fabricating a high density electronic package is disclosed. The package includes a module of laminated semiconductor chips, including spare chip(s) and a supporting substrate with a fixed interconnect pattern. Chip connection pads are provided at a first pad level of the module; one or more pads corresponding to each chip in the module. The module is tested at the first pad level to identify defective chip(s). A spare routing pattern is applied to the module for electrically isolating defective chip(s) and effectively substituting spare chip(s) therefor such that a predetermined pattern of metal interconnect landings on an access surface of the module remains unchanged, as does the supporting substrate.

Intra-module spare routing for high density electronic packages

In a memory system, which includes a dynamic random access memory (DRAM) that has to be precharged before the contents thereof can be selectively read out into a static register, there is provided means for reading the memory contents of the memory cells of a part or the whole of a row of memory cells of the DRAM into the static register while concurrently precharging the DRAM for a subsequent read-out command. This reduces the overall cycle time of the memory array since read-out of the static register can occur during precharge.

Growing community of inventors

Wappingers Falls, NY, United States of America

David J Perlman