Adjust memory directly vs configure DOCP then change settings
Adjust memory directly vs configure DOCP then change settings
This process focuses on what testing involves. Adjusting the case temperature to 16 and keeping the voltage at 1.410v works well. You won't notice much delay, and any extra lag will be recovered through overclocking. I rarely push my B-Die boards above 1.60v with active cooling. Keeping an eye on RAM temperatures is something you're already familiar with...
Check if you've added a temperature sensor to the RAM module. If not, consider installing one for accurate monitoring.
Adjusting DOCP sets a one-time change; you can modify later. Switching back in BIOS from DOCP to DOCP will alter the settings. Keeping it on doesn’t affect manually entered values. I don’t have the exact board, but if it’s similar to mine, changing DOCP may cause BCLK to drop to 99.8 from 100—possibly due to spread spectrum behavior. That means DOCP reduces performance. Therefore, I switched to Manual and input all parameters myself. The 3600C15 RAM should likely run at around 3800MHz with 14-15-14-14 at 1T when GDM is enabled. Voltage needs depend on your other timing constraints. Make sure to configure the correct FCLK in BIOS as well. Ideally, always set it manually in BIOS rather than letting it auto-set, since that might not work as expected from what you’re seeing.
It seems he owns a 3600C15 kit and hasn't noticed anyone except G.Skill advertising them. If it's a TridentZ model, he likely has sensors included. I'm puzzled why top-tier kits aren't marketed to overclockers like the 8-pack b-die kit. The Viper kits are forgivable for their low cost, but you can achieve similar performance with the 3600C15 TridentZ kit at a lower price and with a sensor.
Sure, that makes sense. I usually avoid high voltage work. Just as long as it’s not strictly CPU and memory limited, I’m okay with CL 17 at 4000. There’s enough bandwidth for gaming, so timing isn’t as critical.
I would have liked another Intel machine for RAM overclocking. My 3900X works fine, but it doesn’t feel quite right after hitting the 1:1 limit. I’ve also discovered that breaking the 1:1 ratio reduces stress on the IMC, giving a more accurate picture of the RAM system’s potential. Let me know if this makes sense to others.
Consider the implications after removing the link if memory improvements match Intel's.
I noted this because my newer b-die kits struggle with 1T GDM off compared to older models. When testing 3800C15 at 1T GDM off, it became unstable, especially at 4200C15 where it failed completely. I discovered that at 4200 it wasn't running perfectly in a 1:1 ratio. By adjusting the IF to 1800MHz instead of 1900MHz, the issue disappeared. I'm unsure if an Intel rig would handle 4200C15 at 1T or would need a higher frequency. I opted for a 4400C19 kit, expecting better performance with newer Ryzen 4000 CPUs around 2000-2100MHz. Recently I've been testing it extensively.