The ThrottleStop Guide

Do you have a guide or instructions on how to protect that file from being changed? I'm an administrator on my PC but I can't change the permissions on that file for most of the entries under the Security tab.

Charles

 

you cannot change the permissions because you need to *take ownership* of the files first: right click, properties, security tab, advanced options

you need to change the ownership of "mcupdate_GenuineIntel.dll" from TrustedInstaller to your own admin account

my windows is in another language so i unfortunately cannot help with screenshots, sorry

 

To stop Windows from updating the microcode, you'll need to rename or delete one of the following files (after taking ownership):

C:\Windows\System32\mcupdate_AuthenticAMD.dll -> for AMD CPUs
C:\Windows\System32\mcupdate_GenuineIntel.dll -> for Intel CPUs

This needs to be redone after every big update (1809 -> 1903 etc.).

https://www.win-raid.com/t3355f47-I...code-Repositories-Discussion-18.html#msg86364

---------------------------

https://www.windowscentral.com/how-take-ownership-files-and-folders-windows-10

Edit. See also... https://www.win-raid.com/t5655f47-downgrade-microcode-i-k-z-extreme.html

 

I have a Asus gaming laptop (GL504GM) and my issue is the CPU never goes into anything deeper than C3 sleep state. Other than malwarebytes and some notification icons, there is nothing running.

I'm fairly certain the laptop supports at least C6 and I recall seeing it before.

Any tips to solve this?

 

Your core is in C7, 90%+. Package seems limited to C3. Mine is the same (6700HQ). I think this may be fairly common. Others may have more information on deepest C states for package.
Best,

Joe

 

@chaugh - If it makes you feel any better, my 4th Gen CPU can use Package C6. After I resume from sleep mode, that all changes. Then my CPU gets stuck and can not go any deeper than Package C2.



I pulled all my hair out trying to find out why but I have never found a solution. I compared all of the Windows power plan settings including the hidden ones and I could not find anything. This did not start happening until after I switched to a newer version of Windows 10. Some poorly written driver is likely responsible for this bug. The documentation about what controls the deepest package C state is one of Intel's many well guarded secrets. It might be SSD related. Turning off some C states has been known to give better benchmark scores. That is the only thing I can think of as to why this became someone's good idea. Now users have a very common problem with no solution.

 

Thanks for replying. Once in a while (like on cold boots every few days) I can see it enter C6 but most of the time, C3 is as low it will go.
Whats weird is I made no changes, I never sleep the laptop, always plugged in with the same peripherals and made no config/application changes.

I notice when package only goes to C3, the PKG power is quite high on idle, around 10W (as estimated by TS). When it enters C6, it can go as low as 0.5W.

Do you recommend installing intel® dynamic platform and thermal framework?
On the Asus website for my laptop model, it has it available under recommended drivers. It's from 2018 though (when my laptop mode was first released).

A couple more questions if you dont mind
- when speed shift is enabled, does it matter if speedstep is checked as well?
- I see "enable speed shift when TS starts" - what does this do? My TS is already starting when user logs in.

 

Yee, same problem with my Sager np7870 (clevo N870hp6 chassi) with 7700hq. I've tried everything i found on the forum: clear win10 install, different versions of samsung ssd driver, AHCI driver, iGPU driver, updating IRS, disabling compatibility telemetry (in add to everything recommended in O&O shutup10) - package C still won't go lower than C3 while individual cores are resting at C7 97-98% when idle

 

The package power data is calculated by the CPU internally. It is only an estimation and its only purpose is to control the turbo boost function when the CPU is significantly loaded. This is an interesting number when a CPU is lightly loaded but it is far from 100% accurate. This data was never intended to be used for accurate idle power reporting.

When my CPU package is stuck in C2, I am seeing about 3.5W. When allowed to go into C6, maybe 2.5W. Not a huge difference. If that was a real 1.0W difference, it might be noticeable when running on battery power but not too important when plugged in. Keep an eye on the CPU core temperature. The temperature data is probably a more reliable way to determine how big of a problem this is.

I typically do not recommend installing anything like this. If you think it might help with this C state issue, give it a try. You can always uninstall it later.

When Speed Shift is enabled, the older SpeedStep control method is not being used so it does not seem to matter if SpeedStep is checked or not. I usually leave SpeedStep checked but I do not think this is important.

Newer computers enable Speed Shift automatically in the BIOS. When Speed Shift was first introduced, Microsoft told their customers that if they wanted to use Speed Shift, they would have to install Windows 10. At the time this was true but Speed Shift is a feature of the CPU, not the operating system. This means that Speed Shift can be enabled in any Intel 6th Gen and up CPU that supports this feature. This can be done while using any operating system. Microsoft did not have the time to test Speed Shift on their older operating systems so they decided against offering this feature to Windows 7 and Windows 8 users.

If this option is checked, when you start ThrottleStop, it will check your CPU and it will make sure that Speed Shift is enabled within the CPU. This allows users of older operating systems (Win 7, Win 8) to use Speed Shift. Some laptop manufacturers sold laptops with Windows 10 originally installed but did not enable Speed Shift and they did not give users any way to enable Speed Shift in the BIOS. This ThrottleStop option also allows those owners to use Speed Shift if they want to.

If your BIOS always enables Speed Shift correctly, you do not need to use this feature. Some laptops have problems where the BIOS enables Speed Shift correctly during a clean boot but they forget to enable Speed Shift after you resume from sleep. This ThrottleStop option also takes care of that problem. If ThrottleStop is running and this option is checked, even when you resume from sleep, ThrottleStop will make sure that Speed Shift is enabled.

That is what people like about ThrottleStop, once you learn how to use the program, it will apply settings to your CPU in a very consistent manner. Users of Intel XTU do not always have that feature.

Edit - Tried to do a fair comparison. The first two pics are both with the CPU being limited to Package C2. Similar load and similar idle temps. First pic the power consumption is being reported as 3.1 W, the second pic 1.9 W. Everything seems the same so that does not make too much sense.

The third pic is with Package C6 enabled. Same as the other two pics and now package power is being reported at 2.5 W. The core temps in the third pic are noticeably lower so the true power consumption must be less when package C6 is enabled and power consumption must be less compared to the first two pics. That is why I think core temps are a more reliable way to judge power consumption. Newer CPUs might be more accurate when idle or lightly loaded.

 

Should I keep the core and cache voltage offsets the same? I'm on Skylake-U.

 

Adjust both offset voltages the same for your baseline test. After that, try adjusting the core to a numerically larger value. If baseline is -100 mV for core and cache offsets, keep the cache at -100 mV and set the core to -125 mV or -150 mV and see if it makes any difference to performance or core temperatures. Use something practical like Cinebench R20 for testing purposes.

Laptops have a lot of variables like fan speeds and power limits so it can be very difficult to get consistent results. Remember to go back and forth a few times to prove that your new settings are better. Some users do a single test and call it a day. You need to run several tests to be sure.

 

Thanks for replying! Do you recommend using TS 8.74 if I'm not running 10th gen or should I stick with 8.70.6?

Also, some other questions. What stress testing programs do you recommend? Is it advisable to undervolt the iGPU or should I leave it alone?

 

I like TS 8.70.6 because the .exe is signed by TechPowerUp. There have only been minor updates recently so flip a coin. Use whatever version you like.

I like Cinebench R20, the built in TS Bench test so I can run 1 or 2 Thread tests as well as full load. I still like Prime95 with or without the AVX extensions enabled. If your laptop becomes an inferno running Prime95, set Prime95 to only run 1 or 2 threads to keep the heat down.

I have not done any significant testing while undervolting the iGPU. If you have an Nvidia or AMD GPU, I would leave the iGPU at default voltage. If you do want to try playing with the iGPU voltage, save it for the very end after you are confident that your CPU voltages are 100% stable. I have seen too many people adjust multiple variables at the same time. I prefer adjusting the offset voltages individually.

While trolling the depths of reddit, I came across an interesting quote from a year ago. Unfortunately his name was deleted from his post

He was a real visionary. His vision of the past, present and future are all equally clouded. I am sure some 10th Gen owners are quite happy that I did not listen to this guy and just kept working on ThrottleStop instead.

 

A BIOS update for the Dell G7 7590 has eliminated voltage control. The user has not found a way to go back. He tried using the previous BIOS without any success.

https://forums.tomshardware.com/thr...l-g7-7590-1-11-1-1-12-0.3579943/post-21613665

Perhaps these updates are updating the BIOS and the mcupdate_GenuineIntel.dll file at the same time.

 

Same nasty with this firmware update from Dell as well. Nothing mentioning the serurity advisory... Intel SA-00289 or CVEID: CVE-2019-11157. Nice way trick people (those who know about the problematic fix) into update their firmware... Just hide it and people think they are safe.

https://www.dell.com/support/home/u...ode=wt64a&productcode=g-series-15-7590-laptop

 

What does the BSOD error look like in Event Viewer if I have undervolted too much? I got a random 'Kernel-Power' error today, but I wasn't doing anything CPU or GPU intensive.

 

@rocketvole - When your undervolt is not 100% stable, light load BSOD problems are common. Full load testing is important but putting a single thread load on your CPU is also important. Most users only concentrate on full load testing which is a mistake.

A BSOD is a sign that your CPU needs some more voltage. Do some single thread testing and reduce your under volt.

 

Could you please give us instructions on how to do that?

 

Take Ownership Guide - mcupdate_GenuineIntel.dll

Before making any changes, it is always a good idea to create a System Restore Point just in case something goes wrong.



As you may have heard, due to a security vulnerability, Intel has started blocking CPU voltage control.

https://www.intel.com/content/www/us/en/security-center/advisory/intel-sa-00289.html

Your system manufacturer might release a BIOS update that contains a microcode update but most computers will probably get this microcode update through Windows Update.

If you want to continue using CPU voltage control, make a backup copy of this file
C:\Windows\System32\mcupdate_GenuineIntel.dll

Better yet, follow this guide and take ownership of this file away from Windows. This will put you in control of this file so you can decide if you want to update it or not.

Start by opening the Windows File Explorer and navigate to C:\Windows\System32
Do a search for mcupdate



Right click on this file and select Properties.





Highlight Trusted Installer and then click on the Advanced button.



To the right of where it says Owner, click on the Change button.



Enter the name of your account and then click on the Check Names button.



If this automatically fills in the appropriate user account information, click on OK.



You should see a new Owner for this file. Click on OK.



Highlight TrustedInstaller and press the Edit button.



Now you can edit the Permissions for TrustedInstaller.



I only check Read & execute and Read in the Allow column and I check Write in the Deny column to hopefully prevent anything from changing this file.



You will need to press Yes to agree to a couple of Windows Security prompts.





When you are all done, find the mcupdate_GenuineIntell.dll file, right click on it, select properties and make sure that the changes you made are correct.

 

The guide in my post #11057 is straight forward to follow.

Edit. As well save latest "safe" microcode patches from Microsoft update catalog for later clean OS install.

You'll never know, when you only can get hold of the latest from M$ Update catalog.
http://forum.notebookreview.com/threads/the-throttlestop-guide.531329/page-1104#post-10993126

 

I saw on another forum that 2 users that lost undervolt from a BIOS update got their undervolt back by downgrading BIOS and resetting to default. Just downgrading BIOS alone was not enough for them. The default reset was needed. Anyone who has lost undervolt want to try? I love Throttlestop and want to be able to keep it should a forced/accidental BIOS update happen.

 

Nice guide. Since I am generally blocking all automatic Windows Updates (and have been since before Windows 8 was released) I still had a much older cancer-free version, but I have taken the steps needed to prevent metastasizing.

 

Hi, since a few weeks I get the following error message when TS starts with the system: "error creating service: null".
The message also pops up if I uncheck "Disable and Lock Turbo Power Limits" and then re-enable it (after clicking "OK" obviously).
I tried to whitelist my TS installation folder in Windows Defender as suggested here, but I'm still facing the issue. Any idea ?
I'm with a TongFang GK5CN6Z and i7 8750H proc.

 

@Blebleuh - As you found out, ThrottleStop needs to create that service so the Disable and Lock Turbo Power Limits feature can lock the power limits register.

I am using Windows Defender and I am not having that problem.

What version of Windows 10 are you using? Are you running an Insider version? I have a desktop computer on the mainstream 1909 and I am not having this problem. I will check later today to see the exact version that I am running. I avoid being on the cutting edge of updates so I can avoid problems.

Did you enable the Core Isolation - Memory Integrity feature? This Windows safety feature can block the ThrottleStop Disable and Lock feature.



Try downloading the RwDrv.zip file from Mega again and extract and copy the RwDrv.sys file into your ThrottleStop folder.

https://mega.nz/#!CNNA0SoC!Z2Xi2icwX4d4jzW016dKnKGhVglWmSSPpgiRU7VCG6g

Did you recently add any new software to your computer that might also be using the RwDrv driver?

 

I recently enabled that feature on TS 8.74 on a fresh install, Win 10 pro, have not had any issues. Windows is fully up to date as far as I can tell.

That being said, I have seen errors similar to the poster above. I thought it happened to me after a big windows update, I think my solution was to remove it completely and put the newest version back on and redo the RwDrv.sys steps.

 

I do not even have Core Isolation and Memory Integrity options because I have disabled all available virtualization options in the firmware. Sometimes less is more. A lot less nonsense to deal with. Having a hassle-free experience is definitely worth the minor amount of risk involved with having unnecessary trash turned off. It's terribly unfortunate that most laptops leave end-users with little or no access to useful features and options that are hidden in their firmware.

 

We see the CineBench killing yet another new Dell laptop. Is this some FSB throttling scheme? I haven't seen that before so would appreciate any insight as to what is happening.

- 35.00x 62.365 (top of Throttlestop Screen in red box)
- FID 35x for all cores (first screen shot)
--> cores ~ 100*C & BD Prochot is checked (also prochot 100*C box on right is flagged)

- Then FID drops a few seconds later to about 29x (second screen shots)
--> core temps were closer to 90*C by then

- SpeedShift must be enabled by BIOS (as SST is lit green). I can't see actual EPP on this screen

New 2020 Dell XPS 13 9300 (Icey Lake). You will see ThrottleStop at the 6:40 mark

 

I can't speak for that video but that same thing shows up on my HP Spectre 13" too. It appears to be a false reading as everything else reads the x100Mhz speed. Shows up no matter the temp, even sitting at idle it may say 26 x 6x.xx but is really 2600 Mhz. I haven't had time to investigate, and since I'm thrilled to be able to use TS still I will not complain. Probably my own issue anyway.

 

ThrottleStop 8.70 does not correctly report the BCLK MHz on those CPUs. Not sure if the updated TS 8.74 version works or not. No new hardware to play with to fix this problem. TS works OK on CPUs like the Comet Lake 10710U series.

I enabled virtualization on my 7th Gen Core i but the Memory integrity feature does not seem to work. It lets me turn it on but as soon as I close and go back to that page or reboot, Memory integrity is turned off again.

 

I am using 8.74 and it shows incorrectly, but honestly I would not worry about it. The multi is right and thats all that really matters IMO. We can read the full freq using other means if we really care to. Everything else is working and that is more than we would get without TS

 

I am pretty sure that Intel changed the frequency of one of the timers that ThrottleStop uses. A lot of stuff ThrottleStop does is very hardware dependent. Without new hardware, it is difficult to get everything working perfectly.

@Porter - As you said, the multi is 100% correct so that is all that is really important. You can monitor for throttling with that info.

 

That makes sense @Porter and @unclewebb , especially in context of the video, where the guy is illustrating throttling in process. Thanks for the insight.

8.74 still works great on my Skylake 6300HQ. The SpeedShift EPP function is incredibly convenient to dial up performance (or energy savings).

 

Look in Programs and Features > Turn Windows Features on or off. You may need to install Application Guard or Sandbox from the features console. I'm not 100% positive, but I think those might bear some kind of relation to Core Isolation and Memory Integrity. It would be interesting to know if that does the trick. If it does, then uninstalling them may also do the trick to disable the same "features" for those that do not want them.

 

@Mr. Fox - To get the Memory Integrity feature to enable and stick on my computer, I had to enable a registry value.

Code:

HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\DeviceGuard\Scenarios\HypervisorEnforcedCodeIntegrity

I needed to create a DWORD value called Enable and I set that to 1. This allowed me to enable Memory Integrity. My computer was so secure that ThrottleStop would no longer run at all. Windows was no longer letting the WinRing0 driver load. It also was blocking RWEverything from starting. This means that the Windows Memory Integrity feature does not like the RwDrv.sys driver that ThrottleStop borrowed from RWEverything.

Changing the Enable key back to 0 got ThrottleStop working again. That is enough drama for one day. If anyone hates manually editing the registry but wants to try this, you can run the two different registry keys I created to toggle this feature on or off. If you like ThrottleStop more than you like being safe, you should probably turn this feature off.

Memory_Integrity.zip
https://drive.google.com/open?id=1xcUAP-IVXTMKZ5neoQOI7W_23_WB_64-

 

Silly question, but say we get hit by the microcode or firmware update that blocks undervolting, can we still do some power limit tweaking and multiplier limiting to control temps and throttling at least a little bit? or will everything in TS suddenly stop working?

 

March 10, 2020—KB4540673 (OS Builds 18362.719 and 18363.719)
https://support.microsoft.com/en-ca/help/4540673

I was feeling brave so I decided to install some Microsoft updates today. On my 7th Gen desktop, the microcode version did not change so undervolting continues to work. The Disable and Lock Turbo Power Limits feature also works in this latest Build as long as Memory Integrity is turned off. Not sure about newer 8th, 9th and 10th Gen hardware. It was a large package of goodies so be careful. Wait a few days and let someone else try the medicine first.

@Porter - When under volting gets disabled with a microcode update, adjustment of the turbo ratio limits will also be disabled at the same time. The rest of ThrottleStop should still work OK. If you are using Speed Shift, you can try reducing the maximum Speed Shift multiplier. All of the power limit adjustments should still work OK.

 

I wonder if people using craptastic software bloatware like MSI Dragon Center, Alienware Command Center and XTU will have to create this registry DWORD and set it zero as well. The draconian feces the Redmond Retards are pushing out in a haphazard manner to unsuspecting consumers victims is liable to have undesirable ramifications for more than ThrottleStop. I can see other very useful utilities that use WinRing0 drivers potentially getting screwed up in the process.

 

@Mr. Fox - After I got Memory Integrity working with that registry tweak, Windows decided to lock that option. That made it impossible to turn Memory Integrity back off. I had to manually change Enable in the registry back to 0 to regain control.

A couple of tiny tweaks like this, hidden away in some future Windows Update, will have most users on their knees, begging for mercy.

Windows keeps telling me that the Administrator has disabled this, that or the other. I am the Administrator of my computer. I have to keep finding work arounds so I can use my computer however I want.

 

Man, that really sucks. Maybe a good idea would be to make a Linux version of ThrottleStop and all of us just stop using or supporting Micro$lop Windoze filth altogether. I need to use M$ Office for work every day. If it were not for that, I probably would just stop using Windoze OS X right now. I am seriously considering removing Windoze OS X from my TongFang turdbook and replacing it with Windows 7. It already has a dual-boot setup with Linux, and I am still running Windows 7 on my desktop. I haven't cared about their update rubbish for years and I don't care that they no longer support Windows 7. That's just not relevant to me. Their nonsense with Windoze OS X negates the perceived benefit of "security" that it might offer.

 

How accurate is the power reading in TS? I just noticed my max power was 257.6 w! i7-1065g7, normally I max around 40w during heavy benchmarks/testing

 

Maybe the decimal point is in the wrong spot, or it's mW, LOL.

 

By coincidence I also unleashed updates, including KB4540673 . 6300HQ laptop 6th gen (Dell XPS 9550) with Win 10 Pro somewhat locked down with O&O.

No change in microcode version, undervolting still works, disable and lock turbo power limits still works.

 

The watt numbers are usually consistent without any wild spikes like you saw. Things might get a little screwy when exiting sleep mode with internal timers being randomly reset.

Try turning on the Log File option in ThrottleStop when testing. See if you get an occasional wild and crazy looking value or constant values that are out to lunch. I have never had my hands on a 1065g7. I do not think anything related to power consumption has changed in this CPU family. Are ThrottleStop and HWiNFO usually fairly consistent when both are running together at the same time?

I assume that Linux will also be getting this new and improved microcode update at some point.

 

I fixed that typo for you, bro.

 

I figured out why my cpu package could not go to C6 (stuck on C3) for anyone interested.
I usually keep my laptop plugged into a external monitor, but i have been using it standalone for the last few days.
Without the monitor plugged in, my laptop can go to C6 all the time. Its also much quieter as it idles under 40c.

I'm not sure if its supposed to work like this... but I bet not. When I have time I will do a clean uninstall of the nividia drivers in safe mode and re-install them and see if that makes any difference.

 

Interesting stuff. I have this issue but only when I resume from sleep mode. I am also using an external monitor.

Time to do some testing. With the external monitor unplugged, for almost exactly 60 seconds after resuming, my CPU is stuck and being limited to package C2. After 60 seconds, it is immediately allowed to go into package C6 which is the maximum allowed for this CPU. I have an Nvidia GPU in this laptop but it might be the Intel GPU driver that is doing this trickery. It could also be some sort of Nvidia Optimus bug / feature.

I have a hard drive with a previous version of both drivers. When I have a day to kill, maybe I will go back in time and try to track down when this bug first started. I am pretty sure it used to work correctly even when using an external monitor. Thanks for sharing.

Edit - When I plugged in the external monitor via HDMI with the computer running, it went from package C6 down to package C2 and then after a while, settled in at package C3. Definitely some bugs somewhere.



Edit - After resuming from sleep when connected to an external monitor, I opened up the laptop lid so both screens were on. In the Intel GPU driver control panel, I selected use Single Display and set that to my external monitor. I pressed Apply and instantly, package C6 started working correctly. This leads me to believe that Intel is to blame for these package C state issues.

Unfortunately, every time I resume from Sleep and open my computer display and check the driver, it is back to running in Clone mode which kills the deep package C states. It does not want to remember the Single Display selection that I previously made.

 

@unclewebb I have a Alienware M14xR2 running a 2920XM. This laptop officially supports 3610QM/3630/3720/3740/3820/3840

And unofficially supports 3920XM/3940XM & Sandy bridge mobile series due to HM77 chipset being backwards compatible.


Now the issue...it seems there is a 65W TDP limit issue even with a XM CPU & all the power limits increased.

I remember on my Dell Precision M4800 there's a option that tricks the TDP and let's you OC the 4800MQ but I don't see this option for my 2920XM

Is there any way I can trick the TDP with this CPU? I have the unlocked BIOS as well

 

Do you have the MSR lock disabled and pri-plane maxed out in the unlocked BIOS? These videos are 7 and 8 years old (had to find them in my channel since they were so old) but should be useful in making sure everything got set correctly for the power limits.

 

Yes , I followed those exact guides and it seems there's some sort of 65w TDP limit. This might be due to the 3920XM/2920XM series not officially being supported in M14x/M17x series.

The BIOS settings are exactly the same as M18x. And changing these things does do something since in TS the values change, but it seems there's still a TDP limit I am hitting according to HWinfo

EDIT: What's Non-Turbo Flex support? I notice you have this enabled in R1 but disabled in R2

 

Oh, that's right. There was an issue with the smaller machines having capped power limits. I totally forgot about that. That used to drive Brother @j95 (and others) totally nuts. I don't think anyone ever found a solution to it. If I remember correctly, @svl7 burned a ton of calories trying to fix that and never could find the source of the cancer.