This summer AMD acquired the company MEXT for their AI-powered predictive memory technology to make flash memory such as NVMe SSDs behave more like DRAM. AMD MEXT with EPYC servers allows for flash memory as DRAM tiering and uses a predictive AI engine for handling the tiering in an effective way. There is a Linux driver but this is one of the few offerings in the AMD space that isn't open-source as part of their typical open ecosystem approach. At the AMD EPYC Venice event in Las Vegas, AMD talked up their MEXT software to help with AI-driven memory optimizations and leveraging high capacity NVMe SSDs as an additional tier of memory given today's RAM supply constraints and also the elevated memory prices. AMD MEXT sounds great in principle. Also not too fundamentally different from some Linux memory tiering solutions that exist today, but with an AI engine now to help with data placement and AI driven promoting/demoting between tiers. AMD even talked up their Linux kernel module that "works with any Linux OS." They are promoting MEXT as a feature for EPYC servers. Getting me real excited.... Their published performance claims look promising. MEXT also includes a RAM observability tool. But then came the slides about "MEXT licenses available". Following up with AMD, while they have typically been great at championing an open ecosystem and typically making their software offerings open-source, MEXT is not. At least not right now. MEXT is just available from AMD via their customer/sales channels and isn't open-source nor is it publicly/freely available at this time. Unfortunate but we'll see if that changes in the future or if they are keeping the more restricted access to it for the time being as a sales offering.
AMD Talks Up Their MEXT Software Amid RAM Crisis - Linux Support But Not Open-Source
Full Article
Original Source
Read the full article at Phoronix →KhanList aggregates and links to publicly available news content. We do not host full articles from third-party sources. Always verify important information with original sources.