The document discusses in-memory computing (IMC) technologies and their relation to the von Neumann architecture, highlighting issues such as the memory and power walls. It outlines the purpose and utilization of IMC, emphasizing its benefits in reducing data transportation needs and introducing various physical implementations. Additionally, it addresses common misconceptions about IMC and suggests future work focused on memory state verification and switching variability concerns.

1. 1
In-Memory Computing(IMC)
technologies

2. 2
Content
1. Von Neumann Architecture
2. Von Neumann Architecture’s Problems
3. In-Memory Computing Purpose
4. In-memory Computing Utilization
5. Three common IMC Misjudgments
6. Future work

3. 3
Von Neumann Architecture

4. 4
Von Neumann Architecture’s Problems
 Memory wall

5. 5
 Power wall
Von Neumann Architecture’s Problems

6. 6
Von Neumann Architecture’s Problems
 Machine Learning

7. 7
In-Memory Computing Purpose

8. 8
In-Memory Computing Purpose
 Removing raw data interfaces
 decreasing data transportation power
 decreasing data transportation duration

9. 9
In-memory Computing Utilization
 Physical Attributes such as RRAM, PRAM, MRAM &
FeRAM

10. 10
In-memory Computing Utilization
 array level organization
‐
 peripheral circuitry

11. 11
Three common IMC Misjudgments
 IMC provides competitive value out of the box
 IMC solves current and future performance problems
 IMC allows for a reduction in IT Governance efforts

12. 12
Future work
 verify and correct the memory state
 Switching variability is a major concern