What's a memory booster, and how does it ease SoC bottlenecks
February 28, 2022
By Majeed Ahmad, EDN
What is a memory booster? It’s not another memory but a patented technology that doubles the capacity and bandwidth by employing ultra-tuned compression/decompression accelerators on the data path between the processor and the main memory. The data is compressed when fetched from memory, so the memory access latency is often shorter. That can deliver up to 50% more performance per watt, a crucial advantage for system-on-chips (SoCs) in servers and data centers.
The memory booster IP can be integrated into an SoC device by using the existing on-chip bus protocols. The IP block is placed on the memory access path and is invisible to the operating system and applications.
To read the full article, click here
Related Semiconductor IP
Related Blogs
- Will your multicore SoC hit the memory wall? Will the memory wall hit your SoC? Does it matter?
- Solve SoC Bottlenecks with Smart Local Memory in AI/ML Subsystems
- LPDDR6: A New Standard and Memory Choice for AI Data Center Applications
- LPDDR6: The Next-Generation LPDDR Device Standard and How It Differs from LPDDR5
Latest Blogs
- Satellite communications are no longer as secure as assumed
- Why Hardware Monitoring Needs Infrastructure, Not Just Sensors
- Why Post-Quantum Cryptography Doesn’t Replace Classical Cryptography
- The Silent Guardian of AI Compute - PUFrt Unifies Hardware Security and Memory Repair to Build the Trust Foundation for AI Factories
- Heterogeneous NPU Data Movement Tax: Intel's Own Slides Tell the Story