Intel Presents MemVerge at AI Field Day 1

Introducing Big Memory Computing, MemVerge, and Memory Machine Software

Watch on YouTube
Watch on Vimeo

Charles Fan, Co-Founder and CEO, introduces MemVerge in the context of big memory computing and memory machine software. In today’s computer, applications run in DRAM but store data in storage systems, but this presents challenges for today’s big data applications. MemVerge focuses on big memory computing, allowing applications to run in both DRAM and persistent memory like Intel’s Optane. The MemVerge platform presents a new solution for the cloud era, bringing bigger memory capacity at lower cost. This presentation was sponsored by Intel.

Personnel: Charles Fan

Machine Learning Use Cases for MemVerge Memory Machine Software

Watch on YouTube
Watch on Vimeo

Yue Li, Co-Founder and CTO, presents AI applications for MemVerge Memory Machine software. He begins with training using large memory, showing that DRAM can be combined with Intel Optane Persistent Memory (PMEM) to provide large memory capacity with DRAM-like speed. He uses GraphSAGE as an example of machine learning on a large graph with MemVerge Transparent Memory Service. Next he presents inference with a large model and feature embeddings using Facebook’s DLRM as an example. He also uses image recognition as an example of leveraging MemVerge. Finally he discusses using persistent memory for instant model rollback or recovery.

Personnel: Yue Li