Fact-checked by Grok 2 weeks ago

NVDIMM

A Non-Volatile Dual In-line Memory Module (NVDIMM) is a hybrid computer memory technology that integrates the high-speed access of dynamic random-access memory (DRAM) with non-volatile storage, such as NAND flash memory, to provide byte-addressable persistence during power outages or system failures.^[1] These modules install in standard DIMM slots on the DDR memory bus, enabling direct CPU access while retaining data without continuous power, bridging the performance gap between traditional volatile RAM and slower block-storage devices like SSDs.^[2] NVDIMMs are categorized into three primary types by the JEDEC Solid State Technology Association: NVDIMM-N, NVDIMM-F, and NVDIMM-P, each optimized for different access patterns and use cases.^[3] NVDIMM-N combines DRAM for caching with integrated NAND flash for backup, using a supercapacitor or battery to flush data from DRAM to flash in microseconds during power loss, offering latencies around 100 nanoseconds and capacities up to 32 GB per module.^[2] NVDIMM-F, in contrast, exposes NAND flash directly as a block-addressable device over the DDR bus, bypassing the I/O subsystem for latencies of about 10 microseconds and higher capacities in the terabyte range, functioning similarly to an ultra-fast SSD.^[2] NVDIMM-P merges persistent DRAM with advanced non-volatile media, such as 3D XPoint (discontinued in 2022) or emerging alternatives like ReRAM, providing byte-addressable access to both volatile and non-volatile regions with latencies near 300 nanoseconds, supporting terabyte-scale capacities and enabling new programming models for data-intensive applications.^[2] Development of NVDIMMs began in the early 2010s through industry collaborations, with the Storage Networking Industry Association (SNIA) forming a special interest group in 2014 to promote adoption and interoperability.^[4] JEDEC formalized the standards starting with support for NVDIMM-F hybrid modules announced in 2015, including DDR4 Serial Presence Detect (SPD) codes, followed by JESD245 for the Byte Addressable Energy Backed Interface in NVDIMM-N modules in 2015, and JESD304 for DDR4 NVDIMM-P bus protocols in 2021, ensuring compatibility with DDR4 and emerging DDR5 channels.^[5]^[6]^[7] These standards specify electrical interfaces, power management via backup energy modules (JESD315), and endurance requirements, facilitating widespread integration in enterprise servers from vendors like Intel, Micron, and Samsung.^[8] In enterprise environments, NVDIMMs enhance performance for databases, virtualization, and real-time analytics by reducing data movement overhead and enabling faster crash recovery, with benefits including up to 10x lower latency than NVMe SSDs and improved total cost of ownership through denser, persistent storage.^[9] As of 2025, adoption continues to grow in data centers for AI and high-performance computing, though the discontinuation of Intel's 3D XPoint has shifted focus to alternative persistent media, with challenges including compatibility with legacy systems and evolving software support, such as Linux's DAX (Direct Access) mode.^[10]^[11]

Introduction

Definition and Purpose

NVDIMM, or Non-Volatile Dual In-line Memory Module, is a hybrid form of random-access memory that integrates the high-speed performance of dynamic random-access memory (DRAM) with non-volatility to preserve data during power outages or system shutdowns.^[12] It employs standard DIMM form factors, allowing seamless integration into server and computer motherboards via conventional memory slots. This design enables byte-addressable access, mimicking traditional DRAM while ensuring data retention without external power.^[13] The core purpose of NVDIMM is to overcome the volatility of DRAM, which results in data loss upon power failure, and the comparatively slow access times of non-volatile storage devices such as solid-state drives (SSDs).^[14] By providing persistence at memory-level speeds, NVDIMMs facilitate rapid crash recovery, in-memory data durability, and minimized latency for applications handling large datasets, including databases, analytics, and high-performance computing.^[12] This addresses critical needs in enterprise environments where downtime and data integrity are paramount. Key benefits of NVDIMMs include their ability to support larger effective memory capacities by reducing the frequency of slower disk input/output operations, while maintaining low-latency, high-bandwidth access comparable to DRAM.^[2] In contrast to pure DRAM, which prioritizes speed but sacrifices persistence, or standalone flash storage, which offers durability at the cost of performance, NVDIMMs serve as an intermediary technology that unifies these attributes for enhanced system efficiency.^[15]

History and Development

The development of NVDIMM technology began in the late 2000s, with early concepts emerging to address the growing demands for persistent memory in data-intensive applications such as big data processing and real-time analytics during the 2010s. Initial prototypes were demonstrated by companies like Viking Technology, which unveiled a solid-state drive in DIMM form factor in 2010, laying groundwork for hybrid memory solutions that combined DRAM performance with non-volatile storage. These innovations were spurred by the shift toward cloud computing and enterprise data centers requiring higher bandwidth and lower latencies for handling massive datasets.^[16]^[17] Standardization efforts accelerated in the mid-2010s through the JEDEC Solid State Technology Association, which published specifications for NVDIMM-N and NVDIMM-F in 2015 to enable plug-and-play hybrid modules in DDR4 systems. NVDIMM-N integrates DRAM with NAND flash for backup during power loss, while NVDIMM-F provides direct access to NAND as block storage, both supported by new Serial Presence Detect codes and DIMM signaling for interoperability. The NVDIMM Special Interest Group, under the Storage Networking Industry Association, collaborated with JEDEC to define these standards, facilitating broader adoption in servers. In February 2021, JEDEC released the DDR4 NVDIMM-P bus protocol (JESD304-4.01), extending DDR interfaces to support persistent memory devices with transactional access for enhanced data availability and capacity.^[5]^[3] Key milestones included Intel's integration of 3D XPoint-based Optane persistent memory into NVDIMM-like modules starting in 2019 with second-generation Xeon Scalable processors, offering byte-addressable persistence for in-memory databases and analytics. However, Intel discontinued Optane production in 2022, citing multibillion-dollar losses and a market shift toward technologies like CXL, with last shipments of 100-series DIMMs occurring by December 2023. By 2024, NAND flash-based solutions had become dominant, with NVDIMM-N capturing over 60% of the market revenue due to its balance of DRAM speed and flash persistence in enterprise servers.^[18]^[19]^[20]^[21] As of 2025, trends point toward emerging non-volatile technologies to improve endurance and density in future NVDIMMs, including carbon nanotube-based Nano-RAM (NRAM) announced by Nantero in 2020 for potential integration into persistent memory modules. Early development of NRAM aimed at replacing both DRAM and NAND with a single, high-density non-volatile solution compatible with DIMM form factors. Challenges in earlier designs, such as reliance on batteries for backup, led to their avoidance due to limited lifespan and environmental concerns, prompting widespread adoption of supercapacitors for safer, longer-lasting power delivery up to five years at elevated temperatures.^[22]^[23]^[24]

Types

NVDIMM-N

NVDIMM-N is a hybrid non-volatile dual in-line memory module that integrates volatile dynamic random-access memory (DRAM) for primary data access with NAND flash memory to enable persistence. The DRAM component provides the main storage for active data, offering high-speed, byte-addressable operations, while the NAND flash serves as a backup medium to store a persistent copy of the data. An onboard controller manages data movement between the DRAM and NAND flash, ensuring seamless integration within standard DIMM slots. This architecture allows NVDIMM-N to function like conventional DRAM during normal operation but adds non-volatility through energy-backed mechanisms.^[6]^[12]^[25] In typical operation, the module presents byte-addressable memory to the host system via the DRAM, supporting standard DDR4 protocols for low-latency reads and writes. Upon detection of a power loss event—triggered by signals such as AC mains failure or system shutdown—the controller activates backup power sources, usually supercapacitors, to flush the DRAM contents to the NAND flash. This process preserves all data in the event of an outage, with the flush operation typically completing in approximately one minute, independent of the number of modules installed. On power restoration, the controller reloads the data from NAND flash back to DRAM to resume operations. The backup power plays a critical role in enabling this persistence without interrupting the host system.^[6]^[26]^[27] The NVDIMM-N specification was fully standardized by the Joint Electron Device Engineering Council (JEDEC) in 2015 through the JESD245 Byte Addressable Energy Backed Interface standard, with subsequent revisions such as JESD245E in 2022. This standard defines the host-device interface for DDR4-compatible modules, ensuring interoperability across systems. It builds on the DDR4 SDRAM ecosystem, allowing NVDIMM-N to plug into existing memory channels without requiring specialized hardware changes.^[5]^[6] Performance-wise, NVDIMM-N delivers the high bandwidth and sub-microsecond latencies characteristic of DRAM during active use, making it suitable for latency-sensitive workloads while providing non-volatility for data protection. However, the persistence feature introduces a limitation in the form of the backup flush duration, which can extend to about one minute on power failure, potentially affecting recovery times in rapid restart scenarios. Commercial availability began in 2015 with products from vendors such as Micron, offering capacities like 16GB and 32GB modules in standard 288-pin DDR4 form factors.^[6]^[26]^[28]

NVDIMM-F

NVDIMM-F, or flash-only non-volatile dual in-line memory modules, utilize pure NAND flash memory that is directly mapped into the system's address space, eliminating the need for a DRAM component.^[12] This architecture allows for significantly larger capacities compared to traditional DRAM modules, scaling up to terabytes per single DIMM, making it suitable for handling extensive datasets in server environments.^[2] Unlike the hybrid DRAM-flash design of NVDIMM-N, NVDIMM-F operates entirely on NAND, providing inherent data persistence without requiring external power backup mechanisms.^[15] In operation, NVDIMM-F delivers block-level access akin to solid-state drives (SSDs), but benefits from lower latency due to its direct connection via the DDR memory bus, bypassing slower PCIe or SAS interfaces used in conventional storage.^[29] Data remains intact across power cycles as a fundamental property of the NAND flash, enabling seamless integration into memory slots for enhanced proximity to the CPU.^[12] The operating system treats these modules as block devices, necessitating file system or application-level management for data organization. Standardization efforts for NVDIMM-F have been led by JEDEC, though full interoperability specifications remain ongoing, with the module defined within broader NVDIMM protocols.^[12] Commercial availability began in 2014 from vendors including HGST (now part of Western Digital) and AgigA Tech, targeting enterprise storage applications.^[30]^[31] Performance characteristics of NVDIMM-F include read throughputs exceeding those of traditional SSDs, such as 3-5 GB/s in representative configurations, while latencies are in the range of 10 microseconds—higher than DRAM's nanosecond scale but improved over PCIe-based flash due to the memory channel interface.^[32]^[2] This positions NVDIMM-F for workloads involving large-scale data persistence where capacity outweighs the need for fine-grained access. A key limitation of NVDIMM-F is its lack of byte-addressability, restricting it to block-oriented operations that require larger granularity than DRAM's direct access, thus limiting its use in latency-sensitive caching scenarios.^[12]

NVDIMM-P

NVDIMM-P represents a persistent memory standard that integrates dynamic random-access memory (DRAM) with non-volatile memory (NVM) within a single dual in-line memory module (DIMM), enabling direct byte-addressable access to persistent data alongside traditional block-based operations. The architecture typically employs a hybrid design where DRAM serves as a high-speed cache or buffer, while the underlying NVM—such as 3D XPoint in early implementations—provides data persistence during power loss; data is either mirrored from DRAM to NVM or written directly to the persistent layer for durability. This setup allows systems to treat the module as an extension of main memory, supporting seamless transitions between volatile and non-volatile storage without mechanical components like supercapacitors required in earlier designs.^[2]^[33] In operation, NVDIMM-P utilizes a transactional protocol over the DDR memory interface to manage access, permitting programming models based on load/store instructions that ensure data persistence without the need for explicit full-system flushes or complex backup mechanisms. This enables developers to build applications with true persistent memory semantics, where data modifications are durably committed at the hardware level, reducing overhead compared to software-managed persistence in block devices. The protocol supports interoperability, allowing NVDIMM-P modules to coexist with standard DRAM on the same memory channel, thus facilitating gradual adoption in existing systems. Unlike NVDIMM-N, which relies on power backup for DRAM persistence, NVDIMM-P achieves non-volatility inherently through its NVM integration.^[7]^[2] The standardization of NVDIMM-P was formalized by JEDEC through the release of the DDR4 NVDIMM-P bus protocol (JESD304-4.01) in January 2021, defining the electrical and behavioral specifications for load-reduced DIMMs (LRDIMMs) that incorporate persistent memory while adhering to DDR4 pinouts and topologies. This standard emphasizes scalability for data center environments, enabling larger memory pools by combining the low-latency access of DDR with the endurance of NVM, and paves the way for future DDR5 extensions. Performance characteristics include latencies approaching those of DRAM (on the order of 100 nanoseconds) paired with non-volatility, allowing read/write speeds significantly faster than traditional NAND-based storage; early commercial modules, such as those using Optane, offered capacities up to 512 GB per DIMM, enabling system-level expansions to several terabytes per node.^[3]^[7]^[33] The evolution of NVDIMM-P traces its roots to Intel's Optane persistent memory modules, which demonstrated the hybrid architecture in production systems until Optane's discontinuation in 2022, after which the focus shifted toward alternative NVM technologies like advanced 3D NAND or emerging resistive memory types to sustain the standard's viability. Post-2021 adoptions have emphasized integration with open standards such as Compute Express Link (CXL) for disaggregated memory pools, addressing limitations in capacity and multi-host sharing while maintaining byte-addressable persistence for high-performance computing workloads. Ongoing research highlights NVDIMM-P's role in bridging the gap between DRAM and storage, with implementations showing up to 8 times the capacity of equivalent DRAM configurations without proportional latency penalties in memory-intensive applications.^[34]^[35]^[33]

Technology

Backup Power Mechanisms

Backup power mechanisms in NVDIMMs ensure data persistence by supplying energy during power failures, allowing critical data to be preserved from volatile components to non-volatile storage. Supercapacitors serve as the primary backup power source, delivering short-term energy—typically sufficient for durations of milliseconds to several seconds—to facilitate the flushing of data from DRAM to integrated flash memory. This design evolved from earlier battery-backed units (BBUs), which provided extended hold-up times of up to 72 hours for sustaining volatile memory but required periodic battery replacement and raised environmental concerns due to chemical components.^[36] In implementation, supercapacitors are either integrated directly on the NVDIMM module or provided as tethered external units, such as Micron's PowerGEM or Unigen's PowerGEM, which charge continuously during normal operation from the system's power supply. Upon detection of power loss—often signaled via dedicated hardware pins like SAVE#—the mechanism automatically triggers a controller to initiate the data save process, copying DRAM contents to flash without host intervention. This on-module or tethered approach ensures reliable operation in enterprise environments, where rapid response to outages is essential for NVDIMM-N configurations that blend volatile and non-volatile elements.^[37]^[38]^[39]^[40] Supercapacitors offer key advantages over traditional batteries, including compliance with RoHS directives by avoiding hazardous materials like lead or cadmium, extended operational lifespans exceeding one million charge-discharge cycles, and elimination of maintenance needs associated with battery degradation. These features make them environmentally preferable and suitable for high-reliability applications, providing ample power for the quick flush operations required in hybrid NVDIMM designs. However, their energy storage is inherently limited compared to batteries, constraining backup duration to the time needed for data transfer—often around 100-500 milliseconds for smaller capacities or up to 10 seconds for larger modules like 8 GB—beyond which data preservation cannot be guaranteed. Purely non-volatile NVDIMM types, such as NVDIMM-F, do not require these mechanisms since their flash-based storage retains data without external power.^[38]^[41]^[42] Standards like ACPI 6.0 enable effective power state management for NVDIMMs, incorporating features such as the NVDIMM Firmware Interface Table (NFIT) to describe backup capabilities and coordinate save/restore operations across the system. This integration allows operating systems and firmware to monitor and control power events, ensuring seamless handling of persistence in supported hardware.^[27]

Interfaces and Standards

NVDIMMs integrate with computer systems through the DDR memory bus, with the primary interface based on the DDR4 standard as specified in JEDEC JESD79-4, published in 2012. This enables NVDIMMs to function as drop-in replacements for standard DRAM modules in compatible slots, supporting data rates up to 3200 MT/s in typical configurations.^[43] Specific extensions for non-volatility are defined in JEDEC JESD245 for NVDIMM-N, which outlines the byte-addressable energy-backed interface, and JESD304-4.01 for NVDIMM-P, published in January 2021, which details the bus protocol for persistent memory access.^[6]^[7] Legacy support exists for DDR3 interfaces in earlier implementations, allowing compatibility with pre-DDR4 server platforms.^[44] These interfaces rely on memory-mapped I/O protocols, providing direct CPU access to the module's address space with latencies comparable to volatile DRAM.^[5] For NVDIMM-P modules, the protocol incorporates transactional commands over the DDR bus to manage persistent operations, ensuring seamless interoperability with standard DRAM on shared channels without requiring separate controllers.^[3] This design prioritizes byte-addressable access for high-performance workloads while maintaining electrical and timing compatibility with JEDEC DDR specifications. System compatibility demands server motherboards with dedicated NVDIMM slots and firmware support, such as those paired with Intel Xeon processors starting from the Broadwell-EP family launched in 2016.^[45] Early NVDIMM deployments relied on proprietary interfaces, such as Intel's DDR-T protocol, before standardization by JEDEC in 2015 facilitated broader industry adoption through open DDR-based designs.^[46]^[5] No JEDEC-standardized PCIe variants have been developed for the DIMM form factor, as the architecture emphasizes the DDR bus's inherent low-latency advantages over PCIe for memory-tier applications. As of 2025, DDR5 NVDIMMs are in the sampling and prototyping phase by vendors including Micron and Samsung, offering capacities from 128 GB to 512 GB per module with latencies of approximately 120–150 ns, to exploit the standard's increased bandwidth up to 8800 MT/s for next-generation persistent memory solutions.^[34]

Support

Hardware Compatibility

NVDIMMs require compatible processors to enable their persistent memory features, with primary support from Intel Xeon Scalable processors starting with the first generation (Skylake-SP, introduced in 2017), which includes models like the Xeon Gold 51xx and Platinum 81xx series.^[47] Subsequent generations, such as the second (Cascade Lake, 2019) and third (Ice Lake, 2021), expanded compatibility for both NVDIMM-N and NVDIMM-P variants, particularly Intel's Optane Persistent Memory modules.^[48] AMD EPYC processors provide support beginning with the second generation (Zen 2 architecture, Rome series, 2019), validated for NVDIMM-N solutions in enterprise configurations.^[49] ARM-based servers show emerging but limited native DIMM compatibility as of 2025, often relying on PCIe adapters for persistent memory integration rather than direct slot support.^[50] Motherboards for NVDIMM deployment must feature DDR4-compatible DIMM slots configured for hybrid memory operation, with no need for specialized physical slots beyond standard RDIMM/LRDIMM support, though population rules limit NVDIMMs to one per channel.^[51] BIOS or UEFI firmware updates are essential for module detection, initialization, and power-loss protection management, as the firmware handles memory reference code (MRC) training and non-volatile features.^[40] Adapters may be used in some systems to enable compatibility, particularly for non-standard processor architectures.^[52] In enterprise systems, NVDIMM scaling typically allows 6 to 12 modules per dual-socket configuration, with each module offering 16 GB to 128 GB capacity depending on the type, enabling total persistent memory up to 4 TB per socket or 12 TB in four-socket servers.^[53] This supports byte-addressable access in large-scale deployments, though configurations must balance NVDIMMs with DRAM for optimal performance.^[12] Key vendors include Intel for Optane-based NVDIMM-P modules (discontinued in production after 2022, with shipments continuing until late 2025), Micron (via Crucial branding for NVDIMM-N), Samsung, and Netlist, which provide DDR4-compatible solutions for validated platforms.^[54]^[55]^[49]

Operating System and Software Support

Linux support for NVDIMMs began with kernel version 4.2, released in 2015, introducing the libnvdimm subsystem to manage non-volatile memory devices through the ACPI NFIT (NVDIMM Firmware Interface Table).^[56] This subsystem enables configuration via ND_CONFIG structures for labeling and security features on namespaces.^[57] Direct Access (DAX) support allows byte-addressable access to persistent memory filesystems, bypassing the page cache for low-latency I/O operations. The ndctl utility, built on libndctl, provides command-line tools for namespace creation, region management, and health monitoring of NVDIMMs.^[58] Windows Server introduced native support for JEDEC-compliant NVDIMM-N devices starting with version 2016, using built-in drivers for enumeration and management.^[59] This includes integration with Storage Spaces for pooling NVDIMMs as byte-addressable or block storage, enabling features like direct mapping for databases.^[60] Device discovery relies on ACPI 6.0 specifications, particularly the NFIT table, to describe persistent memory regions to the OS.^[57] FreeBSD provides NVDIMM access through the nvdimm(4) driver, which handles ACPI-enumerated persistent memory devices for block and direct modes.^[61] VMware ESXi supports persistent memory via vPMEM APIs, exposing NVDIMMs as virtual devices in App Direct mode for guest OS byte-addressability or as virtual disks for block access.^[62]^[63] Management tools such as Intel's ipmctl facilitate namespace creation on persistent memory regions, supporting configurations for both memory and App Direct modes.^[64] Security features include data-at-rest encryption using AES-256, managed through passphrases and keys integrated with the Linux keyring or Windows APIs.^[64]^[65] As of 2025, Linux kernel 6.x series continues to enhance NVDIMM-P support through improved libnvdimm handling for pure persistent modes without DRAM buffering, with updates in versions like 6.13 for better firmware interactions. However, integration gaps persist in mobile and edge operating systems, where NVDIMM form factors and power requirements limit adoption compared to server environments.^[15]

Applications

Storage and Caching

NVDIMMs originated as an evolution of battery-backed unit (BBU) caches for RAID controllers, providing non-volatile write caching to protect data during power failures without relying on extended battery life.^[66] In NVDIMM-N configurations, data in the DRAM cache is automatically transferred to onboard flash during power loss via a Cache-to-Flash (CTF) mechanism powered by a small battery or supercapacitor, ensuring indefinite retention until power restoration.^[66] This persistence eliminates the risks associated with traditional BBUs, which are limited to 24-72 hours of hold time, and reduces RAID array rebuild times after crashes by preserving "dirty" (unwritten) cache data for immediate flushing to disks upon reboot, avoiding full parity recalculations.^[66] NVDIMM-F modules function as tier-0 block storage devices, positioned above PCIe SSDs in the storage hierarchy to accelerate I/O-intensive workloads by offering direct memory bus access to NAND flash.^[67] With latencies in the tens of microseconds, NVDIMM-F bypasses the higher overhead of PCIe SSDs (15-100 µs), making it ideal for log-structured filesystems where sequential appends and write-ahead logging benefit from near-DRAM speeds without the endurance limitations of pure DRAM.^[67] For instance, in environments using write-ahead logs, NVDIMM-F serves as a fast persistent tier, reducing the need to stage data on slower SSDs and enabling higher throughput for metadata operations and journaling.^[67] In database systems, NVDIMMs support persistent caching to enhance write performance and durability. For SQL databases such as MySQL, the NV-SQL system leverages NVDIMM as a durable write-back cache for hot pages and redo buffers, reducing SSD write amplification by approximately 50% and boosting online transaction processing (OLTP) throughput by up to 6.5x compared to equivalent DRAM configurations (e.g., 1,831 transactions per second versus 279).^[68] This is achieved through ARIES-inspired recovery protocols with in-update flags, allowing efficient eviction and consistency without full WAL flushes.^[68] Similarly, in Microsoft SQL Server, NVDIMM-N provides a mirrored write-back cache for transaction logs via Storage Spaces, enabling low-latency acknowledgments and faster commit rates.^[69] For NoSQL databases, NVDIMMs accelerate write combining and journaling, as seen in in-memory setups where persistent memory handles tail-of-log buffering to minimize disk I/O.^[70] Operating systems like Linux (via DAX and block translation tables) and Windows Server support these as block devices, facilitating tmpfs-backed persistent caches for temporary database files like MySQL's, yielding sub-microsecond write latencies.^[70] Performance advantages of NVDIMMs in storage and caching stem from their byte-addressable, low-latency access, delivering up to 1.6 million random I/O operations per second (IOPS) at 0.02 ms response time for 4K blocks—roughly 10x the 100-200 IOPS of HDDs and 2-5x the 300,000-500,000 IOPS of high-end PCIe SSDs in random read/write workloads.^[67] These gains are most pronounced in block-access scenarios, where NVDIMM-F's latencies in the tens of microseconds outperform SSDs by up to 10x, though system-level throughput may vary based on bus contention and filesystem overhead.^[67] Enterprise storage vendors have integrated NVDIMMs for caching since 2016, with NetApp adopting the technology through its acquisition of Plexistor in 2017 to enhance ONTAP-based appliances for write buffering and tiering in all-flash arrays.^[71] In NetApp systems, NVDIMMs (as NVRAM) provide non-volatile logging for rapid write acknowledgments, supporting high-IOPS workloads in hybrid and all-flash configurations while reducing latency for metadata and cache operations.^[71]

In-Memory Computing

NVDIMM, particularly the NVDIMM-P variant offering byte-addressable persistence, facilitates in-memory databases by enabling crash-consistent data storage without the overhead of traditional disk-based logging, thereby minimizing recovery times after failures. In SAP HANA, persistent memory serves as non-volatile RAM that integrates DRAM speeds with flash durability to host main data fragments, allowing for rapid read/write operations and near-instantaneous restarts by preserving in-memory states across power cycles.^[72] Similarly, Redis leverages NVDIMM for enhanced persistence in in-memory key-value stores, supporting fast recovery for caching and real-time applications through direct memory mapping that avoids page cache indirection.^[34] In big data analytics, NVDIMM accelerates processing in frameworks like Apache Spark by loading large datasets into persistent memory, which provides DRAM-like latency for iterative queries and reduces data movement costs compared to disk or SSD tiers. This enables Spark workloads to perform low-latency transformations on terabyte-scale data while maintaining persistence for fault tolerance during distributed computations.^[73] For Hadoop ecosystems, NVDIMM supports tiered storage architectures where hot data resides in non-volatile memory, speeding up MapReduce jobs and improving overall cluster efficiency for analytics pipelines.^[73] Virtualization environments benefit from NVDIMM through reduced downtime in operations like VM snapshots and live migrations, as persistent memory allows state capture directly in byte-addressable space without serialization to slower storage. VMware vSphere has supported virtual NVDIMM devices since version 6.7 in 2018, enabling PMem-aware guest OSes to utilize non-volatile regions for resilient workloads, with features like vSphere HA ensuring high availability by preserving VM memory across host failures.^[74] In cloud computing, NVDIMM and persistent memory technologies power memory-optimized instances in hyperscalers like Azure, which supports persistent memory in virtual machines for high-performance computing as of 2023.^[75] For AI and ML workloads, NVDIMMs can protect intermediate data and checkpoints from power loss in GPU-accelerated servers, preventing costly restarts and enabling scaling across multi-node clusters in data-intensive tasks at near-DRAM speeds.^[76] As of 2025, emerging standards like CXL are enabling disaggregated persistent memory for AI applications in cloud and edge environments.^[77]

References

[1]
[PDF] SNIA NVDIMM 1 PAGER vD
A Non-Volatile Dual In-line Memory Module. (NVDIMM) is a “persistent memory technology” that retains data in the event of a power loss or a system shut down. So ...
[2]
Enabling Persistent Memory in the Storage Performance ... - Intel
Jul 25, 2019 · NVDIMM is memory that is nonvolatile and allows for random access. Nonvolatile means that data is kept even without power, so during unexpected ...
[3]
JEDEC Publishes DDR4 NVDIMM-P Bus Protocol Standard
Feb 17, 2021 · JEDEC Publishes DDR4 NVDIMM-P Bus Protocol Standard · Fully compatible with existing DDR channels (Physicals, Electricals, Protocol, Clocking) ...
[4]
SNIA Special Interest Group Formed to Accelerate Awareness and ...
Jan 27, 2014 · A Non-Volatile Dual In-line Memory Module (NVDIMM) is a computer memory module that retains data in the event of an unexpected power loss, ...
[5]
JEDEC reveals NVDIMM standard - Storage - News - HEXUS.net
May 27, 2015 · It is a standard which allows non-volatile memory to make use of the DDR memory bus which has much greater bandwidth potential than conventional storage ...Missing: JESD219 | Show results with:JESD219
[6]
BYTE ADDRESSABLE ENERGY BACKED INTERFACE - JEDEC
This standard specifies the host and device interface for a DDR4 NVDIMM-N, which is a DIMM that achieves non-volatility by copying SDRAM contents into ...Missing: JESD219 JESD245
[7]
DDR4 NVDIMM-P BUS PROTOCOL - JEDEC
An NVDIMM-P device is defined as a LRDIMM memory module which provides host controller access to DRAM and/or other memory devices such as persistent memory. A ...
[8]
Backup Energy Module Standard for NVDIMM Memory Devices (BEM)
Backup Energy Module Standard for NVDIMM Memory Devices (BEM) ... To report potential errors or make suggestions for improvement to a published JEDEC standard, ...
[9]
[PDF] Providing Storage at Memory Speed Using NVDIMMs Sponsored by ...
Apr 14, 2016 · NVDIMMs are persistent memory modules that reside on the DDR DRAM channel, combining volatile DRAM and nonvolatile flash memory. Under normal ...
[10]
[PDF] Persistent Memory, NVM Programming Model, and NVDIMMs
Jun 15, 2017 · NVDIMM modules presented as device links: /dev/pmem0, /dev/pmem1. QEMO support (experimental). XFS-DAX and EXT4-DAX available. 33. Page 34 ...
[11]
What is an NVDIMM (non-volatile dual in-line memory module)? By
Jul 31, 2024 · An NVDIMM (non-volatile dual in-line memory module) is hybrid computer memory that retains data during a service outage.
[12]
Definition of NVDIMM - PCMag
(Non-Volatile DIMM) A family of non-volatile memory standards from the JEDEC Solid State Technology Association (see JEDEC). NVDIMMs use the DIMM DDR4 form ...
[13]
NVDIMM is Changing the Game for Non-Volatile Memory - Nfina
Oct 24, 2025 · NVDIMM, or Non-Volatile Dual In-line Memory Module, is a more advanced form of memory that sits between standard DRAM and non-volatile storage.
[14]
NVDIMMs – A Perfect Blend of Memory and Storage | Synopsys Blog
Sep 9, 2019 · The JEDEC standardized NVDIMMs are: NVDIMM-N: It is the oldest one of all NVDIMMs, where DRAM and Flash are on the same module. DRAM is memory ...
[15]
Viking Modular Solutions Unveils First Solid State Drive in DIMM ...
Aug 27, 2010 · This technology allows users to significantly increase the capacity and storage performance of their existing server, storage or cloud computing ...
[16]
NVDIMM Market Buoyed By Evolving Data Center Demands
Sep 14, 2017 · The shift from the traditional enterprise data center to the cloud is driving an insatiable demand for increased bandwidth and lower latencies.<|control11|><|separator|>
[17]
JEDEC Announces Support for NVDIMM Hybrid Memory Modules
May 26, 2015 · Hybrid modules such as the NVDIMM-N and NVDIMM-F are supported in new DDR4 Serial Presence Detect (SPD) codes to be released later this year ...Missing: 2014 | Show results with:2014
[18]
https://www.servethehome.com/2nd-gen-intel-xeon-scalable-launch-cascade-lake-details-and-analysis/
[19]
Understand the Intel Optane shutdown | TechTarget
Aug 18, 2022 · Intel's upcoming shutdown of Optane follows multibillion-dollar losses. However, some technology benefits came out of the IT giant's effort to improve storage ...<|separator|>
[20]
Intel Optane 100 DIMM Packs Discontinued - ServeTheHome
Oct 16, 2022 · The Intel Optane 100 DIMM 4-packs and 50-packs will have last order dates in June 2023 as the first-gen DCPMMs are discontinued.Missing: NVDIMM | Show results with:NVDIMM
[21]
Non-volatile Dual In-line Memory Module Market Report, 2030
The NVDIMM-N led the market and accounted for the largest revenue share of 60.1% in 2024, driven by its ability to combine DRAM's high-speed performance with ...
[22]
First carbon nanotube NRAM products due in 2020, says Nantero ...
Apr 14, 2020 · The first products are expected from Nantero's licensee Fujitsu Semiconductor in partnership with USJC – a subsidiary of foundry United Microelectronics Corp.Missing: NVDIMM | Show results with:NVDIMM
[23]
Carbon Nanotube NRAM Exudes Excellence in Persistent Memory
Sep 3, 2018 · Nantero NRAM v. NVDIMM. These days, NVDIMMs are becoming popular alternatives for persistent storage on the memory bus. NRAM DIMMs will look ...
[24]
PowerGEM® - Unigen
PowerGEM is a Green Energy Module providing reliable backup power for NVDIMM-N nonvolatile memory modules. Based on supercapacitor technology.
[25]
[PDF] NVDIMM-Ns - SMART Modular Technologies
The diagram below shows the concept of how. NVDIMM-Ns leverage the speed of DRAMs and the persistence of nonvolatile memory to improve system performance.
[26]
Dell EMC NVDIMM-N Persistent Memory User Guide
The backup process takes approximately one minute to complete. This duration is independent of the number of NVDIMM-Ns installed in the server.Missing: flush | Show results with:flush
[27]
[PDF] A Soup-to-Nuts Primer on Using NVDIMM-Ns to Improve Your Storage
Non-Volatile DIMMs, or NVDIMMs, have emerged as a go-to technology for boosting performance for next generation storage platforms. The standardization ...
[28]
[PDF] Micron Introduces Persistent Memory Solution That Combines ...
Nov 10, 2015 · With NVDIMM technology, Micron delivers a powerful persistent memory solution capable of meeting many of today's biggest computing challenges.
[29]
[PDF] Revamping Storage Class Memory With Hardware Automated ...
Jun 27, 2021 · NVDIMM-P such as Optane DC PMM is byte-addressable, but its performance using the app direct mode to support data persistence is yet 6× worse ...
[30]
Micron Launches New NAND-based DIMMs, Intel Announces ...
Nov 16, 2017 · NVDIMM-F has been available since 2014. NVDIMM-P, which isn't supposed to hit market until 2019 at the earliest, actually embeds storage inside ...Missing: vendors | Show results with:vendors
[31]
Non-Volatile Dual In-Line Memory Module (NVDIMM) Market
Apr 3, 2020 · An NVDIMM is termed as a subsystem that incorporates the speed and endurance of a DRAM along with high data retention characteristics similar to ...
[32]
Non-Volatile Dual In-Line Memory Module (NVDIMM) Market
Oct 13, 2025 · In 2024, over 5.2 million NVDIMM units were shipped worldwide, representing a 31% increase compared to 2022. NVDIMMs combine NAND flash and DRAM ...
[33]
Performance characterization of a DRAM-NVM hybrid memory ...
We evaluate the capabilities of this DRAM-NVM hybrid memory system and its impact on High Performance Computing (HPC) applications.<|control11|><|separator|>
[34]
Persistent Memory vs RAM in 2025: CXL & NVDIMM-P Guide
May 29, 2025 · Intel Optane is discontinued, so forward-looking deployments should focus on NVDIMM-P or CXL-based solutions for persistent memory needs. DRAM ...Missing: evolution | Show results with:evolution
[35]
Persistent Memory Research in the Post-Optane Era - ResearchGate
Oct 23, 2023 · Wheres Optane formed a tightly coupled Intelonly ecosystem with all capabilities in a single unified system, CXL is an open multi-vendor ...
[36]
[PDF] DATACENTER MEMORY AND STORAGE - Moor Insights & Strategy
The first use case to market used NVDIMM technology in battery backed up (BBU). DRAM-based DIMMs in the case of a power failure (Figure 21 and further ...
[37]
https://www.micron.com/products/obsolete/obsolete-nvdimm
[38]
[PDF] Product Brief: NVDIMM-N PowerGEM - Unigen
PowerGEM® is a Green Energy Module providing reliable back-up power for NVDIMM-N nonvolatile memory modules. Based on supercapacitor technology, PowerGEM ...
[39]
An NVDIMM Primer (Part 1 of 2) - The SSD Guy
Jul 25, 2017 · In case the name is unfamiliar, NVDIMM stands for “Nonvolatile Dual-Inline Memory Module.” Standard computer memory – DRAM – is inserted into ...
[40]
[PDF] NVDIMM Technical Brief - SNIA.org
Non-Volatile Dual In-line Memory Modules (NVDIMMs) merge two leading technologies: DDR3. DRAM1 and NAND Flash2, to solve the problem of data volatility in ...
[41]
Key differences between supercapacitors and batteries - Eaton
Supercapacitors offer significant benefits over batteries including thermal stability, ultra-long life, and maintenance-free operation.Missing: NVDIMM RoHS
[42]
[PDF] Whole-System Persistence - Microsoft
In our prototype, the microcontroller communicates with Agiga. NVDIMM(s) over an I2C bus. It translates commands sent from the host processor over the serial ...
[43]
Rambus Launches JEDEC-Standard DDR4 NVRCD for Emerging ...
Jun 19, 2017 · Rambus Launches JEDEC-Standard DDR4 NVRCD for Emerging NVDIMM Applications ... NVDIMM-N products and paves the way for emerging NVDIMM-P ...
[44]
https://www.mouser.com/datasheet/2/723/NVDIMM_overview-775686.pdf
[45]
The Uptake Of Broadwell Xeons Begins - The Next Platform
With the Broadwell Xeon launch, HPE created special drivers for Microsoft's Windows Server 2012 R2 for the Micron NVDIMMs and also did BIOS tweaks for ProLiant ...
[46]
NVDIMM-C: A Byte-Addressable Non-Volatile Memory Module for ...
Currently, there are two representative non-volatile dual in-line memory module (NVDIMM) interfaces: a proprietary Intel DDR-T and the JEDEC NVDIMM-P, whic.
[47]
https://support.hpe.com/hpesc/public/docDisplay?docId=a00105399en_us
[48]
Optane NVDIMM backwards compatibility - Intel Community
Sep 4, 2024 · Intel Optane persistent memory is designed for use with the 2nd and 3rd Gen Intel Xeon Scallable Processors family and is enable by a sub-set of processors.
[49]
Netlist: NVMe SSDs and NVvault DDR4 NVDIMM-N Validated on ...
Nov 13, 2019 · In addition, AMD validated Netlist's NVDIMM-N solution on the 2nd Gen EPYC processor and found full functional system and power loss ...
[50]
SMART brings Optane memory to AMD and Arm - Blocks and Files
Apr 13, 2022 · System-on-chip builder SMART Modular Technologies has a PCIe card that it says opens up Optane memory use to AMD, Arm and Nvidia GPU servers.Missing: NVDIMM | Show results with:NVDIMM
[51]
https://www.dell.com/support/manuals/en-us/poweredge-r740xd/nvdimm-n_ug_pub/set-up?guid=guid-933d81b2-c3c6-4aa0-9ca3-43009439b8b4
[52]
NVDIMM and the Linux Kernel - ADMIN Magazine
JEDEC envisages two approaches: NVDIMM-N backs up the content of its DRAM in case of a power outage and restores the data when the voltage is restored. NVDIMM-F ...
[53]
https://support.hpe.com/hpesc/public/docDisplay?docId=a00076327en_us
[54]
Intel schedules the end of its 200-series Optane memory DIMMs
Jul 2, 2024 · Intel officially discontinued its Optane business in mid-2022 and wrote off equipment worth $559 million as part of its broader strategy to ...Missing: 2019-2022 | Show results with:2019-2022
[55]
[PDF] Crucial NVDIMM Product Flyer A4 (EN)
Capable processor and motherboard required to achieve maximum memory speed. ... Crucial NVDIMMs require a compatible hardware platform and BIOS. 2 ...
[56]
How To Install NDCTL and DAXCTL Packages on Fedora - PMem.io
This document applies to Fedora 23 or later running Linux Kernel 4.2 or later. Introduction. The ndctl utility is used to manage persistent memory devices ...Missing: NVDIMM support ND_CONFIG
[57]
LIBNVDIMM: Non-Volatile Devices - The Linux Kernel documentation
The LIBNVDIMM subsystem provides support for PMEM described by platform firmware or a device driver. On ACPI based systems the platform firmware conveys ...Missing: 4.2 ND_CONFIG
[58]
https://docs.pmem.io/ndctl-user-guide/
[59]
Storage-class Memory (NVDIMM-N) Health Management in Windows
Jan 10, 2022 · This article provides system administrators and IT Pros with information about error handling and health management specific to storage-class memory (NVDIMM-N) ...
[60]
Windows Server 2016 persistent memory support supercharges ...
Oct 28, 2016 · You can use NVDIMM-N devices as byte-addressable and block-addressable storage in Windows Server 2016. For example SQL Server log files could be ...Missing: ACPI 6.0
[61]
https://man.freebsd.org/cgi/man.cgi?query=nvdimm&sektion=4
[62]
Using Persistent Memory Devices with ESXi - TechDocs
ESXi supports next generation persistent memory devices, also known as Non-Volatile Memory (NVM) devices. These devices combine performance and speed of ...
[63]
Using Persistent Memory with vSphere - TechDocs
Using vPMem, the memory is exposed to a guest OS as a virtual NVDIMM. This enables the guest OS to use PMem in byte addressable random mode. You must use VM ...
[64]
Security | Version 1.x | IPMCTL User Guide
Dec 13, 2019 · Intel DC persistent memory modules support data-at-rest security by encrypting the data stored in the persistent regions of the DIMM.
[65]
Managing NVDIMM Security - NDCTL User Guide
Mar 10, 2022 · The security management for NVDIMMs is comprised of two parts. The front end uses the Linux key management framework (trusted and encrypted keys ...
[66]
[PDF] Non-Volatile CACHE for Host- Based RAID Controllers - Dell
Traditionally the goal is to maintain cache data for at least 72 hours (Dell guarantees that the BBU will retain data in cache for 24 hours, which is also an ...
[67]
[PDF] The Evolution, Functionality and Benefits of NVDIMMs for Storage ...
Jun 30, 2015 · NVDIMM-N Stores in DRAM, backs up into NAND only on a power loss,. Supercap based, eliminates battery. NVDIMM-F – maps NAND into memory address ...
[68]
[PDF] NV-SQL: Boosting OLTP Performance with Non-Volatile DIMMs
However, unlike scalable PMem, which is typically 3–5× slower than DRAM, NVDIMMs exhibit the same performance characteristics as typical DRAM. Notably, NVDIMMs ...
[69]
Configuring Storage - NVDIMM-N write-back cache - SQL Server
Aug 27, 2024 · This article discusses how to set up a mirrored storage space with a mirrored NVDIMM-N write-back cache as a virtual drive to store the SQL Server transaction ...Missing: NoSQL MySQL tmpfs
[70]
[PDF] Persistent Memory, NVM Programming Model, and NVDIMMs
Traditional Database: Log acceleration by write combining and caching. Enterprise Storage: Tiering, caching, write buffering and meta data storage.Missing: 0 | Show results with:0
[71]
NetApp pockets Plexistor, nonvolatile memory software - Storage Soup
May 25, 2017 · It uses dual-inline memory module (NVDIMM) memory cards, NVMe flash and a spinning disk tier. In late 2016, Plexistor bundled SDM on Supermicro ...
[72]
Persistent Memory - SAP HANA Platform - SAP Help Portal
Persistent memory (NVRAM) in SAP HANA is non-volatile RAM that combines DRAM and Flash, used for main data fragments, and provides fast read/write performance.Missing: Redis | Show results with:Redis
[73]
[PDF] Tiered Object Storage using Persistent Memory - arXiv
Jul 16, 2018 · Quad Core, 32 GB RAM and 512 GB SSD. The Java heap is set to. 4 GB while a separate 5 GB of system memory is reserved as per- sistent memory ...
[74]
[PDF] Persistent Memory Performance on vSphere 6.7 - VMware
Aug 13, 2018 · NVDIMM-N by DELL EMC and HPE: NVDIMM-N is a type of DIMM that contains both DRAM and. NAND-flash modules on the DIMM itself.
[75]
Available on Google Cloud: Intel Optane DC Persistent Memory
Oct 31, 2018 · While the blog post highlights running SAP HANA (SAP's in-memory focused database), you can use them for whatever you want. The persistent part ...
[76]
Non-Volatile Dual In-Line Memory Module (NVDIMM) Market
Dec 25, 2024 · 1. Integration with AI and ML: NVDIMM technology is being optimized to support AI and ML workloads. 2. Expansion of Cloud Services: Cloud ...