Abstract:We experimentally analyze the computational capability of commercial off-the-shelf (COTS) DRAM chips and the robustness of these capabilities under various timing delays between DRAM commands, data patterns, temperature, and voltage levels. We extensively characterize 120 COTS DDR4 chips from two major manufacturers. We highlight four key results of our study. First, COTS DRAM chips are capable of 1) simultaneously activating up to 32 rows (i.e., simultaneous many-row activation), 2) executing a majority of X (MAJX) operation where X>3 (i.e., MAJ5, MAJ7, and MAJ9 operations), and 3) copying a DRAM row (concurrently) to up to 31 other DRAM rows, which we call Multi-RowCopy. Second, storing multiple copies of MAJX's input operands on all simultaneously activated rows drastically increases the success rate (i.e., the percentage of DRAM cells that correctly perform the computation) of the MAJX operation. For example, MAJ3 with 32-row activation (i.e., replicating each MAJ3's input operands 10 times) has a 30.81% higher average success rate than MAJ3 with 4-row activation (i.e., no replication). Third, data pattern affects the success rate of MAJX and Multi-RowCopy operations by 11.52% and 0.07% on average. Fourth, simultaneous many-row activation, MAJX, and Multi-RowCopy operations are highly resilient to temperature and voltage changes, with small success rate variations of at most 2.13% among all tested operations. We believe these empirical results demonstrate the promising potential of using DRAM as a computation substrate. To aid future research and development, we open-source our infrastructure at

What problem does this paper attempt to address?

The paper primarily explores the computational capabilities of Commercial Off-The-Shelf (COTS) DRAM chips and their robustness under different conditions. The core research questions can be summarized as follows: ### Main Research Questions 1. **Possibility of Simultaneous Activation of Multiple Rows**: In DRAM subarrays with a large number of DRAM rows (512-1024 rows), is it possible to robustly achieve the simultaneous activation of more than 4 rows (i.e., multi-row activation)? 2. **Achievable PUD Operations**: By utilizing multi-row activation, what other DRAM-based computations (Processing-Using-DRAM, PUD) can be achieved on COTS DRAM chips? 3. **Robustness of PUD Operations**: How robust is the execution of PUD operations using multi-row activation in COTS DRAM chips? 4. **Methods to Improve Robustness**: How can the robustness of PUD operations be improved? 5. **Impact of Different DRAM Operating Conditions**: What is the impact of various DRAM operating conditions (such as voltage and temperature variations, data pattern dependencies, and different timing delays between DRAM commands) on the robustness of PUD operations? ### Research Findings - COTS DRAM chips can robustly activate up to 32 DRAM rows (i.e., multi-row activation). - By utilizing multi-row activation, COTS DRAM chips can perform some previously unknown PUD operations, such as majority logic operations (MAJX) and multi-row copying (Multi-RowCopy). - The success rate of PUD operations (i.e., the proportion of DRAM cells that reliably and correctly execute PUD operations) varies significantly between different operations. For example, the average success rates for MAJ3, MAJ5, MAJ7, and MAJ9 operations are 99.00%, 79.64%, 33.87%, and 5.91%, respectively. - Input replication is an effective strategy to improve the success rate of PUD operations. For instance, the success rate of the MAJ3 operation increases by 30.81% when activating 32 rows compared to activating only 4 rows. - Data patterns have some impact on the success rates of MAJX and Multi-RowCopy operations, while temperature and voltage variations have a smaller impact on these operations. In summary, the paper demonstrates through experimental analysis the potential of COTS DRAM chips as a computational substrate and provides open-source infrastructure for future research and development.

Simultaneous Many-Row Activation in Off-the-Shelf DRAM Chips: Experimental Characterization and Analysis

SIMDRAM: a framework for bit-serial SIMD processing using DRAM

Understanding the Interactions of Workloads and DRAM Types: A Comprehensive Experimental Study

An Experimental Characterization of Combined RowHammer and RowPress Read Disturbance in Modern DRAM Chips

PULSAR: Simultaneous Many-Row Activation for Reliable and High-Performance Computing in Off-the-Shelf DRAM Chips

DRAMScope: Uncovering DRAM Microarchitecture and Characteristics by Issuing Memory Commands

RowPress: Amplifying Read Disturbance in Modern DRAM Chips

X-ray: Discovering DRAM Internal Structure and Error Characteristics by Issuing Memory Commands

Memory-Centric Computing: Recent Advances in Processing-in-DRAM

CODIC: A Low-Cost Substrate for Enabling Custom In-DRAM Functionalities and Optimizations

Understanding the Security Benefits and Overheads of Emerging Industry Solutions to DRAM Read Disturbance

A Study of Leveraging Memory Level Parallelism for DRAM System on Multi-core/Many-Core Architecture

Experimental demonstration of Single-Level and Multi-Level-Cell RRAM-based In-Memory Computing with up to 16 parallel operations

DRAMA: Commodity DRAM based Content Addressable Memory

Improving DRAM Performance, Reliability, and Security by Rigorously Understanding Intrinsic DRAM Operation

Enabling the Adoption of Processing-in-Memory: Challenges, Mechanisms, Future Research Directions

Mitigation of 1-Row Hammer in BCAT Structures Through Buried Oxide Integration and Investigation of Inter-Cell Disturbances

Multicore DRAM Bank-& Row-Conflict Bomb for Timing Attacks in Mixed-Criticality Systems

Self-Managing DRAM: A Low-Cost Framework for Enabling Autonomous and Efficient in-DRAM Operations

A 3D Stackable 1T1C DRAM: Architecture, Process Integration and Circuit Simulation

RowHammer: Reliability Analysis and Security Implications