AM4 5900x 4000MHz works? Should it be fine or do you need to tweak it?
AM4 5900x 4000MHz works? Should it be fine or do you need to tweak it?
Hello everyone, my current CPU stats are 5900x. The motherboard is MEG-X570-ACE, RAM is 32GB old, G.Skill Trident Z Neo at 3800MHz, and a new 64GB G.Skill Trident Z Royal Elite at 4000MHz with XMP profile 1 enabled. I switched to the 4000MHz version over the 3800MHz. How can I confirm it's stable? Official sources say the Ryzen 5900x doesn<|pad|>, so I only used XMP 1 for the faster memory. Are there any other hidden settings I should adjust to get better performance? Or is it fine as long as it runs without crashes or restarts?
Thank you for your question. The older kit you mentioned is no longer available and has reached the end of its support life. To upgrade to 64GB, you’ll need a compatible upgrade kit with at least 32GB RAM. You can modify the motherboard settings to adjust configurations, but ensure compatibility with your current BIOS and chipset. Lowering the memory clock speed or adjusting timing parameters (from CL18 to CL16 or 15) may help if you encounter stability issues. Always verify compatibility before making changes.
I struggle to remember how to compare CL and clock speed, but with these close numbers it seems the C18 might actually match real-world latency or even be quicker. Keep in mind, CL refers to performance relative to the clock, so a higher CL on faster RAM can match a lower CL on slower RAM. Other timing details might also play a role.
It's simply CL*2000 divided by MHz. In this context, that's accurate. I'm okay with it since it performs better than the typical 10ns of a 3200C16, though it's still not superior to the baseline. However, if you're interested in adjusting, refer to the RAM Tuning Guide by HW Unboxed.
I will call @RONOTHAN## since I only have knowledge and no hands-on experience with DDR4. Your latest RAM is subpar compared to your previous models because the older ones failed and the new ones seem to have some random IC that doesn’t work well with tight subs, especially with TRFC. The 3800C14 is a solid choice. Regarding tuning, I stick to basic primaries and try to clock the RAM as quickly as possible. Decide whether to use sync or desync clock based on frequency differences, using imc and cooling specs for the desync side. Start by tightening the secondary and tertiary rails, then finish with the primaries—never increase voltage on the primaries as they don’t help much. Limitations include cooling, imc power, board specs, and optional voltages like max 1.7V DVRM or 1.3V VOSC; otherwise, don’t rely too much on theory and focus on real performance in games or heavy RAM tasks. Consider testing AIDA64 too, and remember if your CPU gets worse, just sell it and buy another instead of spending hundreds. But if you have a decent sample, aim for stable performance rather than pushing limits.
I'll verify it's acceptable. You mentioned: "hem be because its still faster than the standard theoretical 10ns of a 3200C16, its still not as bad as baseline, it won't matter until we're talking more than a total of 2-5ns of difference." If it exceeds "5ns," it would be a good reason to improve or switch the memory speed. I won't manually overclock or tweak settings. I also didn't run memtest86 since I didn't adjust the configuration manually. Just applied the XMP profile on this 4000MHz CL18 RAM module. Thanks!
I believe there’s always a practical benefit to adjusting memory timings, though it’s probably not enough to justify the effort spent searching for them. It’s frustrating when you think you’ve got them right, only to lose everything because of a crash. Worse still, replacing RAM to achieve those speeds can be costly and time-consuming.