AMD: Where we're going we don't need (as much) DRAM
The House of Zen's Mext team promises smarter SWAP through the power of, you'll never guess: AI
AMD reveals a new AI-powered technology called Mext that could cut enterprise server costs by 83 percent. By loading memory pages in and out of cheaper NVMe storage before the operating system or applications even recognize they need them, this technology could make up to 99 percent of the performance possible with half the DRAM.
The tech, acquired from Mext in June, has already saved customers $5 million in avoided DRAM costs on 145 systems so far. According to former AMD CEO Gary Smerdon, most of a server's cost can be attributed to memory, with 60-80 percent being in use at any given time. However, only a small percentage of those memory pages are actively being used, with 50-80 percent going untouched for extended periods.
Mext aims to offload infrequently accessed pages to flash storage, which is 50 times cheaper and 30 times more power-efficient per bit than DRAM but 500 times slower. To address this speed difference, Mext uses AI models to predict which memory pages will be needed next and proactively loads them back into DRAM, making flash appear as DRAM.
While implementing the technology may result in some performance penalties, AMD believes customers will be willing to compromise for significant DRAM savings.
Written by urgent.news from The Register Science's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
This story
This is one outlet's version. Read the fullest account.
- AMD: Where we're going we don't need (as much) DRAM theregister.com
