Optimize your 2x16GB single rank DRAM module for maximum performance.
Optimize your 2x16GB single rank DRAM module for maximum performance.
Hi, MB : MSI MAG X570 Tomahawk Wifi - BIOS 7C84v181 (Beta) CPU : AMD Ryzen 5800X RAM : Crucial Ballistix 16GB*2 3600MHz CL16 (BL2K16G36C16U4B) #1 I noticed these kits are two single-rank sticks. Do you think they still deliver comparable performance? #2 If I buy the same set, it might cause an IMC overload and be wasteful (plus extra RAM cost and expense), what should I do? #3 I’m attempting a clean overclock to boost my build’s speed. Any step-by-step guide for doing it on a similar setup would help? You mentioned your current O/C is 4000MHz at 16-18-18-38. I’m not very familiar with all the timing details, so how can I improve from here? Should I focus on tuning lower frequencies or adjust timings on the higher one? Also, I ran MemTest (Windows) for two hours without errors so far. AID64 score is around 56ns on some tests. Thanks for your advice!
Great progress achieved! You can observe improvements in latency at lower frequencies, though 4000 seems to be the optimal choice. I haven't adjusted memory timings for AMD yet, but once you've completed your work on Intel, focus on selecting the ideal TREFI and TRFC values. AMD doesn't modify TREFI, so TRFC becomes the next priority. To determine the best TRFC, I perform a binary search. Begin by halving the default value—try 288. If it fails, calculate the midpoint between 288 and 575 (which is 432), then between 288 and 432 (which is 360), and continue this process, always checking the midpoint between the last successful and unsuccessful attempts.
I have comparable RAM models too—I think it's Micron E-Die as well. These modules seem to perform better with lower timings, though they still generate errors during operation. It took a long time to discover the optimal setting they recommended. Remember, I prioritize reliability over performance. My approach was: I used Aida and OCCT to stress-test the RAM and achieve stability. (Started with 1 hour of Aida then extended to 8 hours of OCCT).
Thank you for your feedback. It seems my explanation may have been unclear. My default timing wasn't all set, so when I adjusted the main settings like these: tCL 16, tRCDWR 8, tRCDRD 18, tRP 18, tRAS 38 in Auto, the observed changes were: tRC 58 to 93, tFAW 16 to 42, tRFC 575 to 1100, tPHYRDL 26 to 28. These are the timing adjustments I made based on examples from similar topics. Should I follow the auto values or my own settings? I aim for optimal performance, so a clear methodology would help. If you consider HairlessMonkeyBoy's advice, I should check tRFC 550 instead. Also, I changed the DRAM voltage to 1.40, but my BIOS shows 1.428V. Is there anything else I should adjust to prevent the RAM from failing? My BIOS is set to XML Profile 1, DDR4-4000 FCLK 2000MHz, and the "Improve DRAM latency" option remains disabled.
I will perform a lengthy memory test between each attempt. For better results, consider using 4000MHz 16-16-16-36 settings. Once stability is confirmed after a long memtest64, increase to 4100MHz with an infinite clock speed of 2050MHz.
Hey! Taiphoon mentioned your rank is single rank B Die. It’s good to hear you’re in that position. Just check the screenshot for details before deciding if you can trust it.
What titles are you playing and what configurations do you apply? It seems unclear why a RAM overclock would matter unless you're aiming for maximum performance in competitive settings. (Main concern) If you mainly play single-player games or prefer high detail, you'll likely hit your GPU limits before CPU or memory constraints become issues. (Second point) Should I invest the time if you expect frequent stability tests and adjustments across most of your gaming habits?