Effective clock speed

[i7core]

Hi,

I have a problem with my CPU (8700K), which is that my effective core clock speeds decrease after a while. I have an all core OC set to 4.7GHz and initially when starting the PC, the effective core clocks are 4.7GHz as well when CPU is under load. These are later (like 10 mins later after re-running the test on a cool CPU) decreased to about 4.6GHz.

In my BIOS (ASUS Z390A Prime) I have SpeedStep, Speed Shift and C-States on Auto (I have tried to disable those, but didn’t help). I have maxed out the Power limit settings so that those wont throttle the CPU. I have no thermal issues at all, max core temps during heavy load is like 70C or something.

So, I did not manage to fix the problem via BIOS. However it got fixed by only starting up ThrottleStop version 9.3 without touching anything. I have no idea what it changed and would like to know that. Can anyone guide me through to find out?

 

[unclewebb]

I have no idea what it changed and would like to know that.

I would like to know that too!

I spent days trying to help another user figure out what causes this specific throttling problem that you have. After running many tests, I could not find anything. Without owning any hardware with this specific problem, I finally had to give up trying to figure out why this happens. I rarely give up on issues like this but this one has me stumped.

Run something simple and consistent like Cinebench R20 or R23.

MAXON Cinebench R20.0 Download

CINEBENCH is a real-world cross platform test suite that evaluates your computer's performance capabilities. CINEBENCH is based on MAXON's award-winn

www.techpowerup.com

When your computer has this throttling problem, try running the newer TS 9.4.3. Delete the ThrottleStop.INI configuration file before running ThrottleStop so it is starting with a clean slate. Does this ThrottleStop version solve the throttling problem? Show me a screenshot of ThrottleStop, the TPL and FIVR windows so I can see your settings and also show me the Limit Reasons window. Make sure to exit HWiNFO before doing this test so it does not interfere with the results. Take screenshots while the CPU is fully loaded.

Other users found that TS 9.3 could fix this problem automatically without having to check anything. For TS 9.4 and newer, I got rid of some of the automatic features to make diagnosing these kind of issues easier. For some strange reason, one user told me that TS 9.4 or newer cannot fix this problem no matter how it is setup.

It might be a bug at the CPU level. I think it effects a lot more Intel CPUs than people realize and this bug is present in multiple Core i generations. Most people never notice the throttling. Hang on to TS 9.3. You might forever have to use it to keep your CPU running at full speed.

 

[i7core]

The new version of ThrottleStop did not help with the issue unfortunately. I took screenshots as requested, see attached. I spotted following differences between the 9.4.3 and 9.3 version.

The main window
Version 9.3: Number next to 'Set Multiplier' is 63 T
Version 9.4.3: That number is set to 37

FIVR
Version 9.3: 'Overclock', 'Unlock Adjustable Voltage' are ticked (not in 9.4.3).
Version 9.4.3: 'Ring Down Bin' is ticked (cannot find this option in version 9.3)

TPL
Version 9.3: 'Disable Power Limit Control' unticked
Version 9.4.3: Above is ticked.

 

[unclewebb]

@i7core
Did Limit Reasons show anything lighting up red? Probably not.

The reduced C0% confirms that the CPU is throttling internally. When fully loaded, HWiNFO will report this as reduced Effective Clock.

Try checking the Overclock box in the TS 9.4.3 FIVR window. Push OK after you do that and see if it makes any difference. That is about the only thing that I can see that looks different. The power limits are set to the max so they are fine.

You can also try checking the Speed Shift - EPP box on the main screen and changing EPP from 128 to 0. The EPP number can be edited on the main screen.

 

[kiriakost]

It can not be that easy ... set windows power plan to max performance.

 

[samot]

@i7core , read this thread (posts 8,9,13, and 17), it might help you. Your issue might be the same.

 

[DarkDreams]

I have a similar case with my PC that I never understood but was too lazy to make a post here about. Pretty much the same as OP that after booting I have full performance, but after a while it would just get lower benchmark scores. After checking with HWinfo I found that my C0-Residency would not go to 100% anymore and get stuck around ~97.5% (all cores individually). HWinfo only shows C0, C3 and C7 residency and the remaining 2.5% are nowhere to be seen. I found Throttlestop in the hopes of it showing me more detailed info about the C-States, but lo and behold, just launching it (v9.3) got rid of the problem and even after closing TS my CPU would happily go all out to 100% C0 until I restart or shut down my PC. I simply wrote myself a batch file to start and close TS on boot up and have been running with that.

Edit:
Tested newest Throttlestop beta (v9.4.3) and indeed it does NOT fix the issue in the same way. Even with the same settings as in the old version the problem remained. TS reports C-States 2-10 to be at 0.0% while C0 is around 98%.
Limit reasons is completely empty.
Speedshift EPP enabled/disabled 0/128 makes no difference at all (it is on 128 in my 9.3 version).
Can't use the overclock button since I am on a non K CPU (heck intel)

 

[i7core]

It can not be that easy ... set windows power plan to max performance.

No, that's not it.

@i7core , read this thread (posts 8,9,13, and 17), it might help you. Your issue might be the same.

Seems like that thread is about issues in Windows 11. I have Windows 10.

I have a similar case with my PC that I never understood but was too lazy to make a post here about. Pretty much the same as OP that after booting I have full performance, but after a while it would just get lower benchmark scores. After checking with HWinfo I found that my C0-Residency would not go to 100% anymore and get stuck around ~97.5% (all cores individually). HWinfo only shows C0, C3 and C7 residency and the remaining 2.5% are nowhere to be seen. I found Throttlestop in the hopes of it showing me more detailed info about the C-States, but lo and behold, just launching it (v9.3) got rid of the problem and even after closing TS my CPU would happily go all out to 100% C0 until I restart or shut down my PC. I simply wrote myself a batch file to start and close TS on boot up and have been running with that.

Edit:
Tested newest Throttlestop beta (v9.4.3) and indeed it does NOT fix the issue in the same way. Even with the same settings as in the old version the problem remained. TS reports C-States 2-10 to be at 0.0% while C0 is around 98%.
Limit reasons is completely empty.
Speedshift EPP enabled/disabled 0/128 makes no difference at all (it is on 128 in my 9.3 version).
Can't use the overclock button since I am on a non K CPU (heck intel)

Yep, I have set Task Scheduler to launch TS 9.3 when I start the PC.
I think I'll have to use it all the time.

 

[samot]

Not about Win11, also happens in Win10. But if using another TS version solves it for you, which is great, then the cause for the issue is different.

Seems like that thread is about issues in Windows 11. I have Windows 10.

 

[dnm_TX]

Had similiar experience(laptop) when decided to enable PROCHOT RESPONSE in BIOS. Could be unrelated of course.

 

[Jieqjaw]

Hi, so I currently have this issue.
Had me going mad! Had the issue since about Feb 22 time. Tried everything from stripping my pc to bare bones, flashing the bios, changing all kinds of settings and completely reformatting the Drive reinstalling win 10&11.

I have an 11700k.

Through all my testing I found 3 things…

1) comparing charts of a full performance run at first boot vs a performance drop after being logged on for 10/15m, the only difference I could see was that the effective clock speed drops from 4.6MHz average to 4.5MHz.
2) I found when I did a clean install, I quickly unplugged my internet to stop any updates. I had pre-downloaded all drivers and benchmark software to another drive. I installed them one at time and ran benchmarks between and could not replicate the issue…I tried this over a 2 day period but could not replicate it. As soon as I let windows install it’s auto updates (intel component management) and windows quality updates. The issue reoccurred.
3) As mentioned above in previous threads, installing TS 9.3 fixed the issue.

Conclusion the latest intel updates or windows updates are limiting or affecting C states after a period of being logged on. Hopefully a further update will fix the issue.
for now I have TS 9.3 scheduled to start at logon.

 

[dismission]

Hi, so I currently have this issue.
Had me going mad! Had the issue since about Feb 22 time. Tried everything from stripping my pc to bare bones, flashing the bios, changing all kinds of settings and completely reformatting the Drive reinstalling win 10&11.

I have an 11700k.

Through all my testing I found 3 things…

1) comparing charts of a full performance run at first boot vs a performance drop after being logged on for 10/15m, the only difference I could see was that the effective clock speed drops from 4.6MHz average to 4.5MHz.
2) I found when I did a clean install, I quickly unplugged my internet to stop any updates. I had pre-downloaded all drivers and benchmark software to another drive. I installed them one at time and ran benchmarks between and could not replicate the issue…I tried this over a 2 day period but could not replicate it. As soon as I let windows install it’s auto updates (intel component management) and windows quality updates. The issue reoccurred.
3) As mentioned above in previous threads, installing TS 9.3 fixed the issue.

Conclusion the latest intel updates or windows updates are limiting or affecting C states after a period of being logged on. Hopefully a further update will fix the issue.
for now I have TS 9.3 scheduled to start at logon.

are the windows defender updates (only in windows 11) .... if you replace it the C0 returns to 100%, this is my experience

 

[samot]

are the windows defender updates (only in windows 11) .... if you replace it the C0 returns to 100%, this is my experience

Also happens in Win10.

 

[eviolett]

Not much to add, just saying thanks for posting this since I've been fighting this for about a week (Win10, 9900k), resetting BIOS, reapplying OC step by step, turning every possible knob in the BIOS, reinstalled all the Intel Chipset drivers, etc. I suspected it was a windows update since I was literally working on a new OC and HWinfo was showing Effective clock locked at clock speed, had an automatic update interrupt a stability test, then Effective Clock was 5-150Hz less and C0 State was around 98-99%. I thought maybe I misremembered the Effective Clock, but luckily I take copious notes when OCing and my benchmark scores were lower and temps were 5C lower. Other tools (like OCCT) were showing the same 50-150Hz loss.

I see there might be a solution with Throttle Stop, but I'd rather not install, so if someone somehow finds another fix, I assume you'll be kind enough to post it. In the meantime, I guess I'll wait for MS to fix in a future update <holding breath>

 

[unclewebb]

I guess I'll wait for MS to fix in a future update

This issue has been around for over a year so it is not likely that it will ever be fixed.

but I'd rather not install

ThrottleStop does not make any long term changes to anything. It is a stand alone program. There is nothing to install. When you are done testing, it is easy enough to exit ThrottleStop and delete the ThrottleStop folder. After you reboot, everything will be the same as before.

We are still trying to figure out if this is only a Windows Defender issue or if there is a separate issue that might be a bug at the CPU level. If you can help out, next time you see reduced effective clock, try running ThrottleStop 9.3. For some users, simply running the older version without making any changes will immediately fix this problem. You can exit ThrottleStop after that and the problem does not seem to come back until the next time you boot up. If this works for the problem you are having, I would add ThrottleStop to the Windows startup sequence using the Task Scheduler.

ThrottleStop has an option you can add to the ThrottleStop.INI configuration file.

ExitTime=1

This tells ThrottleStop to exit 1 second after it starts. This should be enough time for ThrottleStop to correct the reduced effective clock throttling problem and then it will quietly exit. In the Options window you can turn off all of the Notification Icons and you can also use the Start Minimized option.

 

[eviolett]

This issue has been around for over a year so it is not likely that it will ever be fixed.

Yeah, I was thinking about it today and it hit me...others reported this a long time ago and when I did my last maintenance (last year) and replaced the thermal paste, I was like "holy crap, this new paste is awesome". It was probably happening back then and I just didn't notice the Effective Clocks until doing a different OC a few weeks ago.

There is nothing to install.

Cool. No problem with these types of apps. Just hate installing and uninstalling, just never seems to get cleaned up.

If you can help out, next time you see reduced effective clock, try running ThrottleStop 9.3

Will do and report back on 9.3. Assuming Intel's XTU doesn't do anything and nobody has been able to replicate the fix with new TS.

Reporting Back: Yep. Merely starting and stopping TS 9.3 "fixes" the issue. One thing I noticed is that when I don't start/stop TS and the C0 doesn't hit 100%, my monitors will go into sleep mode when idle for a long time. After starting/stopping TS, the monitors never sleep. I know in the other thread that MS Defender (or its updates) seems to be the culprit, but wonder why the monitor sleep mode changes with this.

I also tested on my laptop (i7 9750H, Win 10 Pro) and it does not have this issue. Is there something I can help research by comparing the 2?

 

[eviolett]

One thing I noticed is that when I don't start/stop TS and the C0 doesn't hit 100%, my monitors will go into sleep mode when idle for a long time. After starting/stopping TS, the monitors never sleep. I know in the other thread that MS Defender (or its updates) seems to be the culprit, but wonder why the monitor sleep mode changes with this.

Update on the monitors not sleeping. Not true, they still sleep with the TS hack. Just letting you guys know so I don't send someone down another rabbit hole for no reason.

 

[unclewebb]

Merely starting and stopping TS 9.3 "fixes" the issue.

If starting TS 9.3 is able to fix this issue then I think this issue has nothing to do with Windows Defender.

Anything related to Windows Defender might be a completely separate issue.

@samot - Have you ever tried running TS 9.3 when having the reduced C0% issue?

 

[eviolett]

How do I have TS start automatically, but also CLOSE after it loads (like after 5 seconds or something). I thought I saw someone post this somewhere using the INI file, but can't find it. Also, didn't see the option in Windows Task Scheduler.

 

[unclewebb]

Task Scheduler Guide

Under the Conditions tab, make sure none of the boxes are checked.
A gray checked box is still a checked box. Do not assume that a gray checked box is disabled.

 

[samot]

@samot - Have you ever tried running TS 9.3 when having the reduced C0% issue?

Hi @unclewebb ,
Nope, sorry. Haven´t had the need to do that test. I´m still with 9.4.2.
Everyday, 2 or 3 times a day, i do a quick TS bench run to see if the issue has returned. Everything has been fine for the last couple of weeks.
Sometimes i do see the issue returning but unlike as before, if i wait a few minutes everything goes back to normal. On these occasions i´m probably just doing the TS bench run right after Defender has finished updating.

 

[unclewebb]

@samot
Try running Cinebench next time you are testing. When a CPU is not thermal or power limit throttling, Cinebench scores tend to be very consistent. If you are using R23, set Minimum Test Duration to Off so it just does a single run.

I know the Windows Defender Real-time protection feature can trigger this bug but I am not yet sure if it is just a Defender issue or if it is a bug at the hardware level. My testing so far has me thinking that it could be a hardware bug with the system timers used by some 9th, 10th and 11th Gen CPUs. My 10850K scores improve by +900 points when ThrottleStop is used to disable this bug. That is a significant and repeatable difference.

 

[samot]

Probably (most certainly) you don´t recall, but when i began reporting this issue CB was in the mix also. Haven´t used it lately, just the TS bench because it is quicker to check for the issue.

Did a CB run just now, everything is fine. C0% at 100% and ~14200 multi core score on my 11800h. With a reduced C0% i could get as low as 13400, iirc.

 

[unclewebb]

A 6% to 7% improvement in Cinebench scores is typical when this bug is not active. Intel might have fixed this bug in some of their CPUs with a microcode update.

It seems to be a combined bug. On my 10850K, there is definitely something not right with the system timers at the hardware level. When the Windows Defender Real-time protection feature tries to use these system timers, that is when the CPU bugs out and performance takes a significant drop. Not yet sure what percentage of Intel CPUs have this bug.

Thanks for testing.

 

[samot]

@samot - Have you ever tried running TS 9.3 when having the reduced C0% issue?

Hi @unclewebb ,
Nope, sorry. Haven´t had the need to do that test. I´m still with 9.4.2.
Everyday, 2 or 3 times a day, i do a quick TS bench run to see if the issue has returned. Everything has been fine for the last couple of weeks.
Sometimes i do see the issue returning but unlike as before, if i wait a few minutes everything goes back to normal. On these occasions i´m probably just doing the TS bench run right after Defender has finished updating.

Hi unclewebb,

Decided to give other TS versions a try (been using 9.4.2) and i have some questions regarding 9.4.6:

- cpu is a 11800h
- the readme file says: "added feature to increase CPU voltage at 800 MHz for improved stability when undervolting.". Is this value right? I mean, on the FIVR window i have this option: " mV Boost @ 4600mhz".
- mV Boost value isn´t applied
- in the FIVR window the Speed Shift EPP reported value is always 128, even if i switch the Windows balanced power plan between its 3 options - best power efficiency, balanced, best performance. On 9.4.2 the reported value changes accordingly to the power plan.
- unlike on 9.4.2, the Non Turbo Ratio option inside the FIVR window is greyed out.
- and one last thing: in every version from 9.3 up to 9.4.6. any core voltage value set in adaptive mode isn´t applied, and also, the Extra Turbo Voltage option is greyed out.