Underclocking at 9900k despite Oc setting up.huh
Underclocking at 9900k despite Oc setting up.huh
Everything is fine, just like Uncle Webb’s instructions. He’s familiar with Throttlestop.
The first generation Core i7-920 had a maximum multiplier of 21 without C states enabled. Many didn’t understand that turning on C states let it use the 22 multiplier even when lightly loaded. CPU-Z likely didn’t reflect this accurately at the time. The original article explains how enabling C states caused the stable overclock to fail, leading to crashes. Users would reach about 3990 MHz (190 MHz × 21) before instability. Raising C states forced the CPU to switch to the higher multiplier, causing a sudden jump to 4180 MHz and subsequent crashes due to insufficient voltage. This issue persisted in myths about C states being disabled for overclocking. Later Intel addressed it with unlocked multipliers in K-series CPUs. The real cause was voltage limitations at that speed.
It's all relevant; Intel hasn't changed much since the i7 generation. I observed a 22x figure in cpu-z for my 930 (recently upgraded) which was around 190.5, not exactly 22x. I'm currently preparing a Rampage II Extreme for LN2. Previously, I switched from automatic to manual settings, resulting in a 8x-21x gap and 22x reverting back after restarting. It seems the BIOS or manual settings don't automatically disable C-states. I don’t need to disable them when overclocking manually—I’m aware of that. Whether for Rampage II or Maximus X Hero, leaving OC in auto still works. But if I keep OC set to auto while adjusting multipliers, C-states remain active. The CPU throttles inconsistently even at light loads, which bothers me. I’m not saving much by keeping power savings enabled; at least half the threads are gated by the OS when idle anyway.
The Core i7-930 can access a 23 multiplier when the C states are enabled and a single core is active. This too will cause instability if you do not realize that the CPU is switching to this multiplier. Some BIOS options probably do deliberately turn off the C states to prevent the CPU from accessing the maximum multiplier. If you are ever curious, run a ThrottleStop - 1 Thread TS Bench test at default settings. You should see evidence of the 23 multiplier. ThrottleStop reports the CPU multiplier correctly during a light load test. CPU-Z, maybe not. Run the default 133 MHz BCLK to avoid instability at high MHz when doing this test. The 1st Gen CPUs used to do this when the C states were enabled. Check out my 10th Gen screenshot above. Newer CPUs can have high and steady MHz when lightly loaded even if the C states are enabled. @Srius1 That looks much better. No more power limit throttling at 95W.
Thanks! I'll test the throttle stop feature later today. It seems to be working much better now!