Unusual DDR4 3600MHz C16 performance on x570
Unusual DDR4 3600MHz C16 performance on x570
Hey everyone, when I turned on the XMP profile in BIOS, the system emitted three quick beeps and seemed to take about two-thirds of the time to boot. My RAM is running at 2100MHz, but I set it to 3600MHz and 1.35V with XMP still active, but nothing happened. I also tried adjusting it to 3600MHz and 1.36V without success. Recently, I loaded the default optimized settings again and re-enabled XMP, this time setting the voltage to 1.36V. Now it boots fine—Task Manager shows 3600MHz, but Easytune can't detect it and caps at around 3593.52Mhz. HWiNFO lists a value of 1769.8, which equals 18.0 times 99.8MHz. Back in BIOS it now shows 3606.3MHz and 1.39V, which seems to be the real-time reading. What’s happening? Could be a calibration issue or a BIOS update delay?
1.35V could be too inconsistent, and at 1.36V it might lack stability for stress testing. BCLK inaccuracy is a concern, and the board tends to undershoot slightly because the clock generator wasn’t precisely tuned. Lower frequency is preferable over higher one.
They're the base clock, and its frequency is adjusted automatically using the SenseMi algorithm. Reducing the v-core manually can cause noticeable drops in timing, with delays reaching as high as 5 MHz. It's not a set frequency.
Yes, you're interpreting it correctly. The setup seems standard. The BIOS is likely reading the correct frequency and voltage values. A stress test would help verify stability; you can gradually increase load or voltage to see if performance degrades. Raising the voltage isn't recommended unless you're experiencing instability—stick to the specified values unless advised otherwise.
I don't know if BCLK changes dynamically on Ryzen models; it seems fixed since Intel introduced the term instead of FSB. Higher voltage than your setting is typical, though I can't say why manufacturers do it. MEMTEST by HCIDESIGN works well—I open multiple windows matching your CPU threads (or slightly fewer) and run stress tests evenly across them. For instance, with 16GB RAM, an 8-core 16-thread CPU, you could stress-test 15 instances at 1GB each. Reaching 400% without errors should be very reliable.