Comparing performance gains versus timing constraints helps decide the best approach for RAM overclocking.
Comparing performance gains versus timing constraints helps decide the best approach for RAM overclocking.
Hey, on your PC I've been testing RAM speeds and timings. The XMP profile you have is 2400-15-15-15-35. After adjusting just the timing settings, I achieved stable speeds of 2400-12-14-14-28. At 3200Mhz it reached 17-18-18-34. My concerns are: should I keep using the XMP profile or switch to another? If I keep tweaking timings further, could I push past 3200Mhz? Also, is it safe to increase speeds beyond that without worrying about lifespan? I was also thinking about maximizing CL15 speed and then fine-tuning other settings. Lastly, could higher speeds affect the RAM's long-term health? Prime Z370-A i3 8100 3.60 Ghz HyperX Fury Black 2x4GB
If the system remains stable, it makes sense to use the improved configurations. What voltage are you applying? Which tool are you testing for stability? It looks like you're already experiencing increased latency at 3200Mhz. A simple calculation shows that going up to higher frequencies won't necessarily help unless you can bring the CAS down to 16. That means checking CAS levels 12 through 16 and selecting the one that delivers the lowest latency. No, it won't work if you're pushing too high—excess voltage can damage your RAM.
Increase your tRAS level to at least CL+tRCD+2, which equals 17 + 18 + 2 = 37, to prevent issues.
I'm using the standard 1.2V for DDR4 and have kept it steady. For testing, I run AIDA64 and memtest86. Lowering it to 16 causes errors in AIDA64 after a few minutes. I plan to test again tomorrow. It's reassuring. Since I haven't adjusted the voltage before, I'm not seeing any issues now. I wasn't aware of this, but it's helpful to know. Thanks for your assistance. I'll conduct further tests tomorrow, as it's quite late here in Portugal.
Consider the two central points and merge them, such as 13 to 15, then 15 to 30.
It's not focused on stability. tRAS can be adjusted to any level that meets or exceeds tCL+tRCD. A smaller value might cause POST failures or let the IMC take control. The goal is choosing tRAS to maximize performance by preventing premature or delayed closure of active rows. The duration a row stays open before closing is tRCD plus tCL, with an extra buffer of 2 added for optimal results. This approach works well for most users, though the ideal setting varies. Finding the perfect value requires testing and adjustment.
I used my setup to run four passes in memtest86 with this configuration! :3 G.Skill Sniper X DDR4-2666MHz CL13-15-15-30-1T (2x8GB) 16.38GB HyperX Fury DDR4-3066MHz CL14-17-17-34-1T (2x8GB) 16.38GB
Hi @HairlessMonkeyBoy. The testing results show the best performance at 3066-14-17-17-33-1.32V. I haven’t run much stability testing yet, but I plan to do more. My memory has some heatsinking issues, which makes it a bit unstable—I might be able to raise the voltage. It seems unlikely I’ll need to adjust that since I only saw errors in Aida64 during a 15-minute test at 1.3V. Thanks for your assistance!