AMD has detailed its MEXT technology, a predictive memory engine acquired in June. MEXT operates by offloading less frequently accessed memory pages (cold pages) from DRAM to more affordable flash storage. An integrated AI model then predicts when these pages will be needed and prefetches them back into DRAM, aiming to maintain performance levels despite reduced physical DRAM capacity.
According to AMD and MEXT founder Gary Smerdon, the technology can decrease the physical DRAM required in a server by 20% to 75%, depending on the application and performance targets. The operating system still registers the full DRAM capacity. In internal tests with workloads like Synopsys PrimeTime, AMD states it achieved 95% of the performance of a full server while using only 50% of its DRAM capacity with MEXT.
AMD has implemented MEXT internally on 145 servers and plans to expand this to 2,000 servers next year. The company has also announced a partnership with HPE to offer MEXT to customers. The rollout is in its first phase, with some EPYC Venice servers being designated 'MEXT Ready' and equipped with a dedicated NVMe drive or bay for memory offloading. Phase 2 will introduce MEXT licenses and the RAM View application, though a start date has not been specified.
The introduction of MEXT is positioned as a response to increasing DRAM prices, which are affecting enterprise customers beyond just AI workloads. If MEXT delivers on AMD's claims, it could offer a solution for optimizing server memory usage and mitigating the impact of memory shortages. AMD is currently taking a conservative approach to the rollout, focusing on expanding supported workloads based on early results.
✨ This summary was generated by AI from the outlets' reporting listed below. It is not independently verified and may contain errors — check the original sources. How BrevFeed works →
One email each morning: the day's tech stories, clustered across outlets and summarized. No account needed.
One email a day. Unsubscribe in one click, any time.
Spend a few minutes, get the whole day. Every topic's top stories in one hands-free rundown — listen, watch, or read the transcript.
▶ Play today's briefNew every morning, and the back catalogue is archived by date.
AMD is deploying MEXT, a predictive memory engine, to reduce server DRAM requirements by offloading cold memory pages to flash storage using an AI model. AMD claims MEXT can retain 95% of server performance with 50% DRAM capacity in some internal workloads, and plans to expand its internal deployment and partner with HPE for customer availability. This technology addresses rising DRAM prices and potential memory shortages in the enterprise market.