Effective clock readings show discrepancies across OC channels.
Effective clock readings show discrepancies across OC channels.
Just starting here—I'm hoping this is in the right spot and hasn't been covered already. All the discussions I've seen haven't given a clear answer. I have two overclocks: one at 4.6 and another at 4.8. The 4.6 one runs smoothly at 4.6GHz when using prime95, but the 4.8 one fluctuates between 4.5 and 4.8, usually around mid-4.7GHz. I'm not seeing any errors, and none of the power, current, or temperature limits are triggering as reported by hwinfo. Still wondering why they differ.
I noticed it improves with higher voltage (after removing the issues), but I'm nearing my temperature limit at 1.33V core and using llc5—full AVX load (no offset)—reaching 88°C, throttling at 90°C but not showing a drop. It stays at 4.8GHz initially, then drops after some time, though I can't tell if it's because it passed the max test or hitting an unseen limiter.
I know it tracks sleep states in effective clock, which matches what others have mentioned, but why does it spike to 4.8 first and then fall without any noticeable throttling? Also, why doesn't the other OC behave the same way—shouldn't it be sleeping properly?
P.S.: I didn't realize it also works with AVX disabled, where it should handle more power and heat better.
For a second view, leave HWiNFO and run ThrottleStop instead. https://www.techpowerup.com/download/tec...ottlestop/ This method aligns with Intel’s suggested tracking approach and effectively monitors minor shifts in CPU multiplier per thread. At 4.6 or 4.8, the multiplier should read precisely 46.00 or 48.00. Access the Limit Reasons window and observe any red indicators signaling throttling. Close HWiNFO once you’re in that mode to prevent interference with your data. Share some images while the CPU is busy so I can review them. Also, include a screenshot of the TPL window to check your power limits and another of the FIVR window to see your thermal settings. If V-Max Stress or Thermal Velocity Boost are selected, the CPU may slow down when voltage exceeds safe levels or temperature surpasses 70°C. Edit – Your voltage appears slightly high for a 4.8 GHz model. This test used just 10 threads of Prime95 Small FFTs with AVX enabled. It hasn’t been fully validated. Excessive voltage can cause noticeable temperature rises.
Here are the needed images for the 48x OC build, confirming both boxes are active. It appears throttling is occurring, as noted in hwinfo, but the thermal margin seems insufficient for such high loads. Lowering the voltage would have resolved the issue, though higher LLC and reduced voltage might tighten performance. Some users reported running at 5.0GHz at 1.3V, which aligns with expectations for AVX with lower ring counts and tighter LLC settings. MSI mentioned 1.31V for 4.6GHz versus 1.21V for a 9900K, but trends are likely to increase beyond necessity. Stability was maintained at 1.27V with 46x LLC6. The next step could involve unchecking the boxes at 4.7V for improved results, avoiding extreme temperatures. AVX is enabled; otherwise it’s off.
The only limitation I notice is thermal throttling due to the PROCHOT setting at 90°C in the BIOS. When ThrottleStop displays PROCHOT as 90°C in red, it indicates the CPU has throttled. This could have occurred briefly but was sufficient to trigger the sensor. In your AVX screenshot, you're operating near the threshold of throttling. If ThrottleStop shows 48.00 for the CPU multiplier and Limit Reasons is blank, no throttling appears at that moment. Adjusting PROCHOT Offset in the Options window from 10 to 5 may push the CPU into throttling at 95°C instead of 90°C, providing extra temperature margin during testing. Intel has set the thermal threshold at 100°C for reliability. If your CPU demands higher voltage for stability, that's inherent. Limited options exist—Thermal Velocity Boost and V-Max Stress are just toggles, not real indicators. The 9600K likely uses thermal velocity boost, so checking these boxes probably doesn't affect results. My 10th Gen performs well at low voltages below 5.0 GHz but struggles beyond that. The Limit Reasons window may show TVB or V-Max if such throttling is active. Overall, everything seems normal as long as the multiplier stays steady at 48.00 in ThrottleStop. Temporarily reduce turbo limits and run a stress test; you should see improvements in Limit Reasons. Red boxes mean throttling is happening, while yellow records past events.
Thanks for the feedback! It seems this OC should be fine, no issues with the limits or multiplier changes. At higher loads it stays consistent, and the PROCHOT will handle the final tests well—it won’t reach those extreme levels during regular use.
It seems like the system was behaving oddly before, with unexpected throttling that only showed up in clock data. After restarting and saving the settings, everything returned to normal, but it now runs at a stable 4.8ghz effective clock, which is better than previous versions. I’m puzzled about why saving and reloading the same configuration made such a difference, especially since the settings looked identical. On both updates, the stress tests kept showing yellow limits (PL1 and EDP Other), but clearing them removed those warnings permanently.
CPU often activates throttling indicators during initial boot or after power-saving modes like sleep or hibernation. These signals appear before the BIOS can establish CPU power limits, which is why clearing such data is recommended before testing. Exercise caution with ThrottleStop; modifying BIOS settings may cause existing configurations to reset. It’s wise to remove the ThrottleStop.INI file after any BIOS changes to prevent overriding saved values. I prefer using Cinebench R20 for consistent results. If you manage to replicate throttling issues, try running R20 again without this problem to compare performance. You’ve sparked my interest now. I’ll explore further with HWiNFO.
I managed to reproduce the issue and it didn't match my expectations (better to wait before making assumptions when working on autonomous systems). I'm hoping you can shed some light on what's happening. I began the system at its exact settings, which reached full 4.8ghz last night. Verified that Throttlestop.ini remained deleted, let it warm up while I enjoyed my coffee, launched hwinfo and Cinebench R20, and it continued throttling without any adjustments or even entering BIOS mode. Then I closed hwinfo and restarted Throttlestop (without the .ini file), ran the test again and it performed better. After that, I disabled limit reasons, left Throttlestop running, opened hwinfo just to confirm what I thought—throttling wasn't active after starting Throttlestop. I deleted Throttlestop.ini, restarted, and still didn't see throttling. Rebooted and it didn't affect BIOS either. I ran hwinfo and Cinebench R20 again, throttling occurred once more. Small FFT tests in Prime95 Avx also showed throttling, but not at the expected levels (no throttles reported in hwinfo except for effective clock). This time I kept hwinfo active, stopped the test, restarted Throttlestop without the .ini, and reran the same checks. The FFTs still throttled, but it hasn't reached the point where it should. It seems the problem might be related to Throttlestop itself, even though I didn't modify any settings in that file. However, Hwinfo also confirmed this, which is concerning. Unfortunately, Throttlestop isn't identifying the cause before testing begins, possibly because of prior assumptions about BIOS behavior. It appears it's connected to Throttlestop, not just a configuration issue.
ThrottleStop begins its operation by examining CPU registers for discrepancies. The BIOS is intended to standardize core values, yet occasional BIOS flaws can cause inconsistencies. If the system detects any irregularities, it quietly resolves them without alerting users. This automatic correction works regardless of whether an INI configuration file is present. Using HWiNFO alone, did you check for throttling indicators? There are many factors that trigger throttling, and Intel processors typically report issues promptly. If nothing appears, consider using CPU-Z to review the status. Open the About section, click Save Report (.TXT), and generate a report post-boot before ThrottleStop. Repeat with another run after applying the fix. If you share these reports later, it might help identify any changes. I just tested by disabling hyperthreading on a 10-core processor; I ran Cinebench R20 at 4800 MHz. My setup used a 9600K clock and DDR-3600 memory at 16-16-16 MHz. Performance gain was minimal compared to the 9th or 10th generation chips.
Still observing no throttle indicators in hwinfo prior to applying Throttlestop. Photos were captured during a Cinebench R20 test where throttling occurred before using Throttlestop. I also executed prime95 afterward to recheck, but nothing changed: CPU-Z logs before Throttlestop, during a Cinebench R20 run while throttled: CHALK1.txt logs from CPU-Z after Throttlestop, and during a Cinebench R20 run without throttling: CHALK2.txt. I plan to review these later to spot any variations, though it seems the comparison will likely be at a superficial level—things appear fairly similar overall, which makes sense given the complexity. The fake 9600k model caught my attention. I’d have expected a noticeable increase in performance due to stable operation at lower voltages, which would typically raise temperatures. My RAM operates at a higher clock speed, while mine uses tighter timings (3200MHz, CL14-14-14). It’s possible the boost comes from a larger cache at 4500MHz, though I’m confident it’s another factor not clearly explained. Since I manually set the ring to 43x, the core voltage stays high even at idle—using more power instead of lowering it. This could explain the higher wattage consumption. Given how much I use my computer daily, balancing performance and battery life is crucial, especially at $0.27 per kWh. This brings up another question: why does hwinfo display C7 state residency, while the BIOS limits it to C3 for faster wake-up? I noticed stuttering during scrolling after pauses in low load when full C-state was active. This issue likely began with game-related stutters, prompting BIOS auto-settings to push the CPU up to 1.38V at 46x—something I was unaware of for a year while running Folding@home. The consensus now leans toward 1.35V as safe, possibly up to 1.4V if temperatures are low. My experience before this optimization was better with lower voltages, which explains the higher power draw. This might also relate to why 48x requires more voltage. Edit: the only meaningful difference I grasped was the altered ring count during throttling, though it still idles at lower voltage and draws less power when running normally after Throttlestop—suggesting the BIOS adjustment is affecting auto settings without obvious interference.