AMD seems to be working on an auto memory overclocking feature that will greatly simplify the tedious task if it manages to pull it off right. A patent submitted to the USPTO describes the implementation of this simple-looking but convoluted utility. The flowcharts indicate a straightforward approach, one that will require quite a bit of patience on the users’ part. The memory frequency will be bumped up gradually till the system becomes unstable, followed by a comprehensive stability test to ensure smooth operation.
Alongside or/and separately, the memory timings will be tightened (CAS, RCD, RP, RAS, RC, CR?) to determine the lowest stable figures. Then it will be stability tested along with the determined overclocked frequency to generate a profile.
There are two submitted models. One that overclocks only the primary memory timings, and another that tests the subtimings as well. This rather complex algorithm first tests the memory frequency for the highest stable point along with one or more primary timings. The combination is then stability tested separately and in tandem. Finally, the subtimings are tested for the lowest stable figures with the frequency and primary timings set to the figures determined in the previous steps.
Three overclocked profiles are then generated, one with just the frequency, the second with the frequency and primary timings, and the third with the subtimings boosted as well.
It’s unclear whether AMD will release the AMO feature alongside the Ryzen 7000 processors in the next couple of months but considering the increased focus on memory overclocking, including RAMP (Ryzen Accelerated Memory Profile), exclusive DDR5 support, and faster Infinity Fabric Interconnects, I’d say it’s only a matter of time.
