Increase CPU cache speed through overclocking.
Increase CPU cache speed through overclocking.
Well LLC is recognized for maintaining a high idle voltage. I haven’t tested LLC7 yet to compare results with LLC6. Currently, I’m using FFmpeg with uncore at 46 to observe the outcome. This might seem unnecessary since I’m uncertain if faster memory improves performance.
I understand the issue remains at elevated idle voltage. The setup likely uses a fixed voltage offset, so it should stay near 1V rather than 1.4V. It seems the system was configured for a static voltage of 1.45V—LLC level 8 should include some droop even during idle. Since I don’t use FFMpeg much, I can’t say much about Uncore’s impact. Benchmarks such as Y Cruncher, SuperPi, and others gain significantly from higher uncore performance.
LLC8 should not have a vdroop since it's the top level on my board. I believe they're referring to transients rather than just dips and spikes. Thanks for those benchmarks—I'll definitely give them another try. Also, I never set a static voltage; it's always been adaptive.
ASUS doesn’t use LLC backwards. I’m more familiar with MSI and ASRock boards where Level 1 avoids VDroop and Level 9 avoids LLC. You’d want to skip Level 8 (Asus), as it’s the loudest choice and needs a much higher VCore for stability. Levels 2 or 3 seem to be the best options.
They mentioned what I said about level 8 having no VDroop and that idle and load voltages match. The issue with an LLC circuit reaching its limit like on level 8 is that it creates a very noisy VCore rail. This applies to the X570 Godlike (likely) rather than the Maximus XIII Hero, but the same applies. Additionally, MSI actually reverses VDroop—level 1 stays flat while level 9 has significant VDroop.