AMD has filed a patent for a program that performs automatic memory overclocking to simplify enthusiasts' lives. The app can automatically test overclocked memory modules for stability and locally store memory overclocking profiles unique to a particular system.
Dynamic Random Access Memory (DRAM) overclocking has existed for decades. At first, companies like Corsair cherry-picked memory ICs that could safely work at high frequencies and/or with low latencies. It then built modules that were guaranteed to work at specific frequencies, timings and voltages but required those settings to be entered manually into BIOS. To simplify things, Intel introduced its eXtreme Memory Profiles (XMP) — settings stored in a module's serial presence detect (SPD) chip and can be activated on supporting platforms — in 2007, and virtually all enthusiast-grade memory currently comes with XMP profiles.
But guaranteed/recommended settings and XMP profile settings tend to be relatively conservative to maximize compatibility with a broad array of systems. Meanwhile, each PC is unique, so diehard enthusiasts tend to determine maximum data transfer rates and the lowest latencies their machines are capable of themselves. However, this is time-consuming and requires knowledge, experience, and skills.
AMD's automatic memory overclocking technology attempts to give average performance enthusiasts the ability to tune their memory modules and CPUs for maximum performance like a pro, but without extensive manual testing. In addition, AMD's automatic memory overclocking should also allow owners of inexpensive memory modules with JEDEC standard settings in SPDs to overclock them to their limits. Performance-minded enthusiasts who tend to use the best RAM modules featuring XMP 3.0 profiles will also be able to maximize their performance to levels that their builds can handle without doing loads of manual work.
"Settings for memory modules can be configured according to vendor profiles or user input specifications," AMD's patent US20220155982 reads. "These profiles are often generated and tested using system configurations differing from a user system. Additionally, user input specifications may be restricted using margins determined and tested using these differing system configurations."
AMD's automatic memory overclocking technology entails determining an overclocked memory frequency setting, passing the memory stability test, determining the most aggressive timing settings for passing the said test, and generating a profile that includes the overclocked memory frequency and latency settings. The resulting profile will eventually be loaded by the program. Interestingly, the patent does not touch upon memory voltages, which means that modules must maintain SPD voltages (i.e., voltages from SPD profiles), or voltages must be set manually in the UEFI BIOS or using a third-party app.
The patent was published on May 19, 2022, so we can only guess as to when AMD's program will come to fruition. However, we could safely speculate that AMD will make it available once it launches the next-generation Ryzen 7000-Series processors.
The examination process will take a few more years, so if a patent is eventually granted, by then we will have all forgotten about it.
Isn't the Ryzen Dram calc basically just looking up values from a list of 'known good' settings for Zen 2?