Memory Trends,Recent Results and research gaps area discussed by HP
Why Memristor-based Memories?
– RRAM with good scaling
• 5 nanometer junctions demonstrated
– Some memristor material systems are nonlinear
• Enables use in crossbars
– Crossbars
• Amortizes current steering transistors over more cells
• Achieve higher bit densities and lower cost than 1T1R
Memristor Research Needs
– Writing techniques for crossbar arrays
• Single bit vs. multiple bits per row, etc.
– Crossbar sensing schemes
– Impact of dead cells and mitigation techniques
– Error coding tailored for memristive crossbars
– 3D techniques
• Full 3D crossbar, 1-D arrays of 2D crossbars, etc.
– Wear leveling techniques for crossbars
Nanostores will use memristors to Co-locate Data (Memory) and Processing
Nanostore Research Needs
– Programming models for “hierarchy of computes”
– Applications built using said models
Photonics and Memristive memories
– Photonics and memristive memories have lots of potential
– Much research is still needed:
• Accurate prediction of future memory technologies
• Efficient circuit techniques for memristive crossbars
• Fault tolerance and error correction for crossbars
• Programming models and applications for “hierarchy of computes”
• Programming models utilizing distributed non-synchronous checkpointing
• Cost of near-term photonics
• Demonstration of robust DWDM CMOS-compatible photonics
• Technologies for constructing integrated high-radix switches
If you liked this article, please give it a quick review on ycombinator or StumbleUpon. Thanks
