Shift from DRAM to NAND AMD Acquires Memory Technology Startup Mext

From | Translated by AI 2 min Reading Time

Related Vendors

To address memory bottlenecks in data centers, AMD has acquired the American memory startup Mext. The technology developed by Mext relies on offloading rarely used data to NAND flash, thereby keeping DRAM costs low.

AMD has acquired the California-based startup Mext. The storage tiering technology developed by the company is designed to offload infrequently used data to NAND flash, thereby reducing the demand for DRAM in data centers.(Image: Dall-E / AI-generated)
AMD has acquired the California-based startup Mext. The storage tiering technology developed by the company is designed to offload infrequently used data to NAND flash, thereby reducing the demand for DRAM in data centers.
(Image: Dall-E / AI-generated)

AMD has announced the acquisition of the startup Mext. The company develops memory optimization technologies intended primarily for use in data centers. The purchase price was not disclosed. According to available reports, AMD’s stock rose significantly following the announcement.

Behind the acquisition lies a growing problem in IT infrastructure: When it comes to large AI models, data analysis, virtualization, and high-performance computing, the processing power of CPUs and GPUs alone is no longer sufficient. Increasingly, available storage capacity is becoming the limiting factor.

Mext is working on a storage tiering technology designed to allow operating systems and applications to use NAND flash as if it were RAM. NAND flash is significantly cheaper than DRAM, but slower. The technology is designed to help use expensive DRAM more efficiently and offload less frequently accessed data to less expensive storage media.

Prediction Rather Than Mere Outsourcing

At the heart of the technology is the so-called Predictive Memory Engine. It continuously analyzes access patterns and uses AI models to predict which data will be needed next. This data can then be moved back from flash memory to DRAM in a timely manner, before applications access it.

The goal is to provide applications with more usable memory without requiring operators to equip their systems with additional DRAM to the same extent. For cloud providers and enterprises, this could mean lower infrastructure costs and better utilization of existing systems. However, it remains to be seen in practical applications whether these benefits can be achieved on a large scale.

For AMD, Mext complements its existing portfolio of processors, accelerators, networking components, and software. The company competes in the AI chip market with Nvidia, among others, and is seeking to offer data center operators a broader range of solutions. In addition to the technology, AMD is also gaining a team with expertise in memory architectures, infrastructure software, and large-scale computing systems.

The most important task now is integration. Mexts' approach must be integrated into AMD's platforms and validated under real-world workloads. Only then will it become clear whether the technology can significantly reduce memory bottlenecks in AI and data center environments. (sg)

Subscribe to the newsletter now

Don't Miss out on Our Best Content

By clicking on „Subscribe to Newsletter“ I agree to the processing and use of my data according to the consent form (please expand for details) and accept the Terms of Use. For more information, please see our Privacy Policy. The consent declaration relates, among other things, to the sending of editorial newsletters by email and to data matching for marketing purposes with selected advertising partners (e.g., LinkedIn, Google, Meta)

Unfold for details of your consent