MEXT
AI overview
Sign in and the AI will write an overview from our coverage.
Headlines · 2
- AMD Details Venice EPYC Roadmap and MEXT Memory Tech
At its In Depth technology day, AMD detailed the Zen 6-based sixth-generation EPYC 9006 roadmap, dividing Venice into four platforms for performance, enterprise, HPC and AI-head-node workloads. Venice SP7 will ship first. AMD says SP8 is planned for the first half of 2027, while Venice-X and Verano are planned for the second half. AMD also presented MEXT, software that uses predictive models and NVMe flash to move data between DRAM and flash and reduce dependence on DRAM. The performance and savings figures cited in the report came from official tests or AMD.
36氪 人工智能 · 🔥 0 - AMD unveils MEXT to reduce server DRAM use with AI and flash
AMD introduced MEXT, a memory-management solution that uses an AI prediction engine to offload cold memory pages to flash and return them to DRAM before use. AMD says it can reduce physical DRAM requirements by 20% to 75%, with internal tests reporting 95% of full-server performance at 50% DRAM capacity in Synopsys PrimeTime, though Tom’s Hardware says the undisclosed test methodology limits verification. AMD has deployed the technology on 145 internal servers and announced an HPE partnership, while commercial licensing remains part of an unscheduled second phase.
Tom’s Hardware · 🔥 0
Experience and discussion from the community
Share my MEXT experienceAsk about MEXT
Nobody has shared their experience with MEXT yet.