Memory module with data buffering

Patent No. US11093417 (titled "Memory module with data buffering") on Nov 25, 2019. The application was issued on Aug 17, 2021.

What is this patent about?

’417 is related to the field of high-density memory modules, specifically focusing on architectures that allow a memory module to support more physical memory ranks than a standard memory controller is designed to address. It addresses the technical challenges of signal integrity, electrical loading, and timing synchronization that arise when increasing the number of memory devices on a single printed circuit board.

The underlying idea behind ’417 is to decouple the physical memory devices from the system memory bus by using an intelligent intermediary layer that manages both control signals and data paths. By implementing a logic-controlled buffer, the invention allows the module to present a simplified electrical load to the controller while internally managing a more complex array of memory ranks. This insight allows for load isolation, where the high capacitive load of multiple memory chips is hidden from the memory controller, enabling higher speeds and greater capacities.

The claims of ’417 focus on a memory module architecture that utilizes logic to receive input chip-select signals and output registered address and control signals to specific N-bit wide ranks. A critical aspect of the independent claims is the inclusion of specialized circuitry positioned between the memory bus and the data pins of the memory devices. This circuitry is controlled by data buffer signals and is specifically configured to operate with an overall CAS latency that is intentionally higher than the actual operational latency of the individual memory chips.

In practice, the invention works by registering data transfers through the intermediary circuitry for a specific time delay, typically at least one clock cycle. This extra cycle provides the necessary timing budget for the logic to decode commands and route data to the correct physical rank without violating the strict timing requirements of the high-speed memory bus. The data buffer control signals dynamically enable and disable data paths, ensuring that only the active rank is electrically visible to the bus during a read or write operation.

This approach differs from prior solutions by moving beyond simple signal buffering to a comprehensive rank multiplication strategy that is transparent to the host system. While traditional modules are limited by the number of chip-select lines provided by the controller, ’417 uses its internal logic and registered data paths to simulate a virtual memory configuration. This allows the use of lower-density, cost-effective memory chips to achieve the performance and capacity of much more expensive, higher-density components while maintaining superior signal integrity.

How does this patent fit in bigger picture?

Technical Landscape

In the mid-2000s when ’417 was filed, memory subsystem architectures were typically implemented using a direct electrical interface between a memory controller and a limited number of memory ranks mounted on a module. At a time when systems commonly relied on standard registered DIMM architectures, hardware constraints related to signal integrity, load fan-out, and bus capacitance made increasing memory density beyond two ranks per slot non-trivial. Because each additional memory device added an electrical load to the high-speed data and address buses, engineering constraints often forced a compromise between total system memory capacity and the operational frequency of the memory bus.

Prosecution Position

The disclosed invention represents a technical advancement through an architectural shift that decouples the memory bus from the individual loads of the memory devices. By integrating logic for registered address and control signals with specialized circuitry for data path isolation, the system enables the use of a higher number of N-bit wide ranks than standard controllers typically support. This configuration achieves a technical effect where the memory controller perceives a reduced electrical load, allowing for increased memory density without degrading signal integrity. Furthermore, the architecture enables an intentional registration delay in data transfers, allowing the module to operate at high data rates while managing the timing constraints inherent in high-density, multi-rank memory configurations.

Claims

The patent contains a total of 15 claims, with claim 1 serving as the sole independent claim. This independent claim focuses on a memory module architecture featuring a printed circuit board, control logic, multiple ranks of memory devices, and specialized circuitry designed to manage data transfers between a memory bus and the memory devices by introducing a specific time delay that makes the module's overall latency greater than the internal latency of the individual memory devices. The dependent claims generally serve to further define the physical and operational characteristics of the module, including load isolation, specific rank and device configurations, clock signal synchronization, data path management, and the reporting of latency values to a memory controller.

Key Claim Terms New

Definitions of key terms used in the patent claims.

Term (Source)Support for SpecificationInterpretation
Actual operational CAS latency
(Claim 1)
Data transfers through the circuitry are registered for an amount of time delay such that the overall CAS latency of the memory module is greater than an actual operational CAS latency of each of the memory devices.The inherent Column Address Strobe latency setting at which the individual DRAM chips are functioning, which is distinct from and lower than the module's total latency.
Data buffer control signals
(Claim 1)
The logic is further configurable to output data buffer control signals in response to the read or write memory command. The circuitry is configurable to transfer the burst of N-bit wide data signals between the N-bit wide memory bus and the memory devices in the one of the plurality of N-bit wide ranks in response to the data buffer control signals.Control signals generated by the module logic in response to memory commands to manage the timing and direction of data flow through the module's isolation circuitry.
N-bit wide ranks
(Claim 1)
The DRAM devices of a memory module are generally arranged as ranks or rows of memory, each rank of memory generally having a bit width. For example, a memory module in which each rank of the memory module is 64 bits wide is described as having an “×64” organization. The memory devices are arranged in a plurality of N-bit wide ranks.A grouping of memory devices on the module that collectively match the data width of the memory bus (N-bits), allowing the rank to receive or output a full burst of data in response to a single command.
Overall CAS latency
(Claim 1)
Data transfers through the circuitry are registered for an additional amount of time delay such that the overall CAS latency of the memory module is greater than an actual operational CAS latency of each of the memory devices. The circuitry is configurable to transfer the burst of N-bit wide data signals... in accordance with an overall CAS latency of the memory module.The total Column Address Strobe latency of the entire memory module as seen by the memory controller, which includes the internal latency of the DRAM devices plus additional delays introduced by the module's circuitry.
Registered chip select signals
(Claim 1)
The ranks of a memory module are selected or activated by address and command signals that are received from the processor. Examples of such address and command signals include, but are not limited to, rank-select signals, also called chip-select signals. The set of registered address and control signals includes a plurality of registered chip select signals corresponding to respective ones of the plurality of input chip select signals.Chip select signals that have been processed and output by a logic device (such as a register) in response to input chip select signals, used to specifically activate one N-bit wide rank of memory devices while keeping others inactive.

Litigation Cases New

US Latest litigation cases involving this patent.

Case NumberFiling DateTitle
2:23-cv-00628Dec 22, 2023Netlist, Inc. v. Micron Technology, Inc. et al

Patent Family

Patent Family

File Wrapper

The dossier documents provide a comprehensive record of the patent's prosecution history - including filings, correspondence, and decisions made by patent offices - and are crucial for understanding the patent's legal journey and any challenges it may have faced during examination.

  • Get instant alerts for new documents

US11093417

Application Number
US16695020A
Filing Date
Nov 25, 2019
Publication Date
Aug 17, 2021
External Links
Slate, USPTO , Google Patents