Evaluating the benefits of DDR4-3733 overclocking on an AMD 5900X and checking for micro stutters is worthwhile.
Evaluating the benefits of DDR4-3733 overclocking on an AMD 5900X and checking for micro stutters is worthwhile.
I have adjusted my RAM to 3733 14-15-15-15 and am checking if it’s fully stable. I’ve added small tweaks to the dimming and SOC settings to avoid errors during testing. The tools I’ve tried include DRAM calculator membench, Prime 95, AIDA64, and others. I’m unsure if any can detect micro stutters or determine long-term value. Should I keep at 3733 or return to a lower frequency for better stability? To assess stability, I need ways to spot micro stutters or frame drops. Running the 3600 OC at 1.420V stable would let me revert if 1.48V feels risky.
Your results are solid and your system runs smoothly, which means you've reached a consistent overclock. If you're focused on performance metrics, try running benchmark tests like Unigine Superposition to check for improvements.
Your response times at 3600 were consistent. Focus should always be on speed rather than data rate. You can compare performance using AIDA64. Most users overlook how timing depends on frequency. Please accept this if you know, but from a latency standpoint, 3200Mhz CL16 matches 3600Mhz CL18. The 1.48V level is safe, and 1.5V reaches its maximum safe daily setting, which is the highest XMP voltage I use.
I noticed inconsistent latency spikes around those times. It might be worth investigating potential micro-stutters or frame rate fluctuations during quick tests.
The process involves checking stability through multiple tests. Create a bootable USB for memtest86 and run it through several test cycles. This can be time-consuming but helps identify issues on the IMC. For extra assurance, use Prime95 to stress all threads with FFT sizes from 512 to 4096, dedicating 75% of your RAM and spending about 15 minutes per size. Monitor WHEA values in HWInfo64 or Event Viewer. Cinebench often fails when RAM is unstable, so consider repeating the test. If games run smoothly and no WHEA errors appear, you're likely safe. I haven't encountered RAM-related stuttering before, especially without other clear problems.
I wasn't considering RAM issues leading to frame rate drops or stuttering. I'd heard somewhere that exceeding 3600 involves a lot of work and can cause tiny glitches, which was especially noticeable at 3800.
I understand this might be a new concept for you. It seems like it could stem from someone who isn't fully aware of what they're discussing. As long as your MCLK, FCLK, and UCLK are in sync, you should be okay. However, if you're experiencing stuttering above 3600Mhz and it's not just a placebo, consider raising the SOC voltage up to a maximum of 1.2V or reducing your RAM back to 3600Mhz. Performance differences usually aren't significant here, but you likely already know that.
The ideal maximum voltage for SOC during regular use is typically around 1.1 to 1.2 volts. Your current reading of 1.09 suggests you're operating near the safe limit—keep monitoring for stability.
As you gain more experience, concerns about the issue lessen. I’m not very experienced with the Ryzen IMC, so I can’t give precise advice. My knowledge is more aligned with Intel IMC behavior. Still, once you consider the Primary timings, Tref and Trfc have the biggest impact on performance. You can’t adjust Tref with a Ryzen unit. With my B-die kit, which seems to be a smaller package, I can set Trfc to 267 at 3600Mhz and 296 at 4000Mhz. Doing some calculations suggests around 280 at 3733Mhz, though I’m not sure if the Ryzen IMC will influence this.