AMD Ryzen Memory Tweaking & Overclocking Guide 151

AMD Ryzen Memory Tweaking & Overclocking Guide

Theory: Relations Between Timings »

Theory: Sequence for Tuning RAM / SOC

The most important and fundamental step is to be able to start the system at the desired frequency. To do so, we need to manually set specific settings in the BIOS:
  • XMP memory profile (some motherboard manufacturers name it differently, but the function does not change)
  • The frequency for the RAM (this means the desired frequency of the user)
  • The BCLK frequency (if the motherboard supports such a setting in the BIOS)
  • The timings (as recommended by the calculator)
  • The voltage for SOC and DRAM (as recommended by the calculator)
  • The procODT + RTT values (NOM, WR, and PARK).
For more details, the DRAM Calculator for Ryzen instructional video can be found here.

To begin, enter the parameters from the calculator for procODT + RTT (NOM, WR, and PARK), trying the recommended and alternative values until the system has a minimum number of errors. For this, use the TM5 0.12 test package with the basic preset. This will of course not get rid of all the errors that may appear. To do so, we will also need to follow through on the next step.

Next, find the optimal voltage for the DRAM and SOC, which will further minimize the number of errors the system has. First, select the voltage for the SOC, then for the DRAM (the calculator will tell you the available range for each). Sometimes, a digital multimeter or voltage monitoring software can help in determining if the value set in the BIOS is the actual amount being delivered. The reasons why these might not match are many and beyond the scope of this document. But since it may have an impact, it may be prudent to verify the voltage with a program such as HWInfo if stability cannot be established no matter which values are set. To check for errors, use the TM5 0.12 test package with the basic preset.

In half the cases, you can get a fully stable system at this stage. If the TM5 0.12 test package does not find errors, increase the range of test programs to check for stability. These can include Linx, HCI, Karhu, and other programs. If none of those programs find an error, continue to the next step—the debugging.

As indicated in the figure below, debugging has as its main goal to change certain timings.

You should check on how each timing adjustment effects stability of the system. I do not recommend changing all the timings mentioned at once. Patience is key at this stage. Do one variable at a time. If the tested timing does not improve the situation, reset it to its prior value and check on the next timing on the list.

Any further advice and all additional steps beyond this point are meant for more experienced users familiar with overclocking (generally, more than a month of experience). I will primarily present these steps as flow-diagrams with the primary parameter indicated above these images.

Tuning the CAD_BUS

Adjustment of Additional Voltages

Each diagram lists a set of parameters to be tried at each stage. I created these diagrams in a way that has the higher-priority settings that can improve stability checked first. Of course, you can take a different route as there are no clear rules and laws on which settings may help stabilize your system at this point.

Timings: Limits and the Benefits

Recently, I've been noticing that the optimization of memory has become a fanatical hobby for some people, with outcomes that may please them and other users. In fact, running the system at extra-low timings will not guarantee maximum performance. According to the results of my research, this is far from the case; sometimes, low timings can even be the reason behind "freezes".

I extensively tested the effect of timings on gaming performance, using memory with different ranks (SR 3466 and DR 3200). The processor was overclocked to 4.2 GHz. The main reason to overclock the processor for these tests is to turn memory into a bottleneck for the entire system. In both cases, I compared the XMP preset with results achieved by manually tuning each timing. The timing I changed has been made bold. The priority of timings for alternate tuning is chosen at my own discretion, from the most significant to the less important ones.

As you may notice, underestimating some timings, such as tWTRS, can adversely affect system performance. It can also be concluded that the XMP profile, although called "eXtreme", is not extreme. On average, XMP profile tuning allows you to gain up to 14% average FPS. Also, another minus of the XMP profile is the stability of the system in most cases. Focus on the recommendations of the calculator and you should have no problems choosing the optimal timings for maximum gaming performance.
Next Page »Theory: Relations Between Timings