The ThrottleStop Guide

@illuzn - If the Microsoft Surface is not using the deep package C states then I call BS! It sounds like they do not know what they are doing.

When I first got my hands on that Lenovo C930, with Avast antivirus installed and not much else, it could easily spend over 80% in package C8.



Package C9 and package C10 are disabled. That part of ThrottleStop does work but I have seen very few screenshots that show any activity in either of those C states. Manufacturers are not using these.

Installing the Adobe Creative Cloud bloatware dropped over 20% off of the package C8 residency time. Sometimes worse. Even when not using the Cloud bloatware, Adobe has dropped piles of .exe files into the startup sequence that are constantly running. Programmers that do stuff like that make me cringe.

If you open up a command prompt with Admin privileges, type in this command.

powercfg -qh >C:\test.txt

That command will create a file in your C drive called test.txt. Upload this file somewhere and send me a message so I can have a look. There might be a deep sleep option that is set differently. That command queries the power settings including the 101 hidden power settings. It might take me a few days to do some testing and a thorough comparison but I am as interested in solving this mystery as you are. Intel has created some low power C states but many manufacturers do not bother fully using these. In the early days of C states about 10 years ago there were some stability problems when overclocking to the moon but I think those problems are all in the past.

If you have anything checked in the Options window that requires battery monitoring then this option will remain checked. Post a screenshot of your Options window if you need help. If you have AC Profile or Battery Profile checked then ThrottleStop is going to have to continue monitoring the status of your battery.

 

Indeed, AC and Battery proflies are checked, so it's normal Thanks

 

I do see profound increase in PKG power on SST 0 and my 6700HQ can go to c6 state only. Maybe some other PC with deeper C state unlocked can run SST 0 w/o increasing PKG power. Your temps looks like Papusan's temps.

 

TS 8.70 with Cinebench R20, solid 3.9GHz throughout no throttle thx to TS

ROG GL703GS - 8575H, 32GB @2666, GTX 1070, NVMe & SSDH, W10 Pro - 1809, stock paste​

Q-6

 

Keep in mind that the package power shown by all software is only an estimation. It is not actual power consumption. On the 8th Gen 8550U, it seemed pretty consistent regardless of CPU speed or VID voltage. On earlier CPUs, this estimate used the VID voltage as part of its calculation. More VID voltage resulted in higher reported power consumption.

That is why you should look at the CPU core temperatures when making comparisons. More actual power being used by the CPU will show up as higher temps. In the screenshot I posted, core temps were close to identical even though VID voltage was twice as high and the MHz were three times as high.

 

Arf, my cpu don't go to C7 anymore, it stays at C3...
Do you guys know where it could come from? Pretty sure it's Throttlestop related as I didn't changed anything else
Thanks

EDIT : After rebbot, it goes back to C7 normally... But even with process explorer, I couldn't find the guilty. Any idea?

 

You need to uncheck BDPROCHOT and check off Speedshift. I dunno how you managed the CPU clocks to work that way w/o Speedstep.

Yeah I know. With speedstep enabled I can achieve that but the downside is slight loss in responsiveness and super fast transition to higher clock states on Speedshift.

 

There's no throttling, nor is the GPU under stress with this test, so BDPROCHOT is not an issue, if it was it would be flagged in Limit Reasons. By default Speed Shift is enabled by W10 which is superior to Speedstep with an Speed Shift EEP of 0 set by W10 offering full performance of the CPU.

Q-6

 

You use Windows speedshift? I found it was not quick enough. Windows speedshift driver works based on power plans and TS works almost independently. I use Balanced power plan and use TS to control speedshift.
PROCHOT sensor has been triggered when you reached 94C.

 

Unless you disable Speed Shift in W10 it will conflict with TS, and if W10 is setting Speed Shift EPP to zero in Best Performance there's no need to use TS to control Speed Shift. Generally I leave both W10 and ROG performance settings on max, unless I explicitly want the notebook to be quiet by using the TDP down feature.

ASUS has it set up that 95C is the point of thermal throttle on the GL703GS, as long as it's below the CPU can deliver a full 3.9GHz across all cores. It's just a good CPU and holds a solid undervolt, although I don't generally run it at the limit for obvious reasons as I use the GL703GS professionally.

Q-6

 

TS can bypass Win 10 SST values.

 

My understanding is the way W10 is set up now it will conflict with TS with Speed Shift checked, I don't know the fine details, equally it's not going to run any faster in this case as EPP is already at 0 and CPU holds max frequency. Therefore no need to use that aspect of TS, 8750H locked so 3.9GHz (6 cores) is all you get if the notebook is capable enough.​

Q-6

 

8.70 won't conflict with SST of Windows anymore. If its works for you then its good!

 

I agree with that 100%. I am not sure how things used to be but when using Windows 10 - 1809, the Windows Best Performance setting sets the EPP value to 0 so the CPU can run at its maximum speed. That is all anyone can ask for. There is no need to check the Speed Shift EPP box in ThrottleStop if Windows is finally taking care of this and setting EPP appropriately.

To find out what Windows 10 is doing on your laptop, do not check the Speed Shift EPP option in ThrottleStop, open up the FIVR window so you can monitor the EPP value in the top right corner and then move the Windows power slider back and forth and see what EPP values Windows is using.

On the 8th Gen Lenovo laptop I tested, it changes EPP to 84 when the Windows slider is set in the middle. That has proven to be a good compromise, similar to the traditional Balanced profile so once again, no need to adjust this in ThrottleStop. If Windows is doing something right, no need to override the EPP setting.

It is only necessary to clear the BD PROCHOT box if a laptop is actually using this throttling method. In his case, the Limit Reasons screen shows no evidence of throttling. BD PROCHOT is not being used so the state of the ThrotleStop BD PROCHOT box does not matter. Some laptops use BD PROCHOT excessively for throttling purposes while other laptops do not use this at all or perhaps only use this throttling method when there is a legit reason to use it.

I still smile every time I hear an 8750H owner say that, especially when ThrottleStop is being used to achieve that goal.

 

when i play BF1, if i use EPP at 0, and fivr radio limits as normal ( 8750H at 3.9 ), i got high temps on cpu, reaching 96° on max and 86° average. I prefer lowering the epp to 96 with fivr at 3.6 Mhz, my temps are on 84° max on 2 cores, average 72° on BF1 and it runs smooth.

Edit : Undervolt done for both test EPP ( -151.4 cache and core, igpu and unslice 48.8 )

 

Whatever works for you. You might get temps a bit lower with a little judicious undervolting.
Best,

J

 

I believe this is can be configured by your device OEM. When plugged in "best performance" gives me an EPP of 0 but when on battery the maximum EPP is "84" on "best performance" and 174 on "recommended".

There's probably a hidden setting in the power profile somewhere.

 

I've not personally tested v1809 for speedshift values on my PC because I don't have official support for SST so I use TS to engage or disengage EPP_HWP. v1903 obeys TS's EPP value setting w/o a fuss!
I disable BDPROCHOT while using Performance profile and only on battery I enable it.

 

didn't mention it, but already done, thx ( -151.4 cache and core, igpu and unslice 48.8 ), just edit my post.

 

I want to report that Throttlestop is working normal without any problems on latest win10 1903.
But Memory Integrity Check must be disabled.
BR

 

Why is it that in Throttlestop, it always says 128 next to Speed Shift - EPP? I change the min and max values in TPL and that affects my FID. Is it normal that I set min to 14 and max to 14 and my FID becomes 14? I see people here post about how 1 or 0 is the max performance, yet when I set it there my computer acts as if it is throttled. 255 min and 255 max is where I see it go to 3400 MHZ.
(XPS 15 9560 i7-7700HQ, Intel Thermal Management is off, undervolted, turbo off, speedstep on)

 

You can edit the EPP value in ThrottleStop. I guess it always says 128 because you have not changed it yet.

The values in the TPL window control the FID or multiplier. The EPP value is the one that goes from 0 to 255 with 0 being maximum performance and 255 being minimum performance. Both of these values work together. If you set the Minimum and Maximum multipliers sky high in the TPL window then the EPP value will be ignored.

Yes, that is normal. Tell me what you are trying to accomplish and I will tell you how everything should be setup.

On 8th Gen CPUs running Windows 10, I prefer to let Windows manage the EPP value as long as it is setting EPP correctly.

Why? Is this another laptop from Dell where they forgot to include an adequate heatsink and fan?

 

I am confused on how to change the EPP value. I have seen some people talk about it here, but cannot see those options available on my ThrottleStop.

 

https://i.imgur.com/sWhMzIp.png

 

On the main screen, change the EPP value from 128 to whatever you like. It really is that simple. That 128 number that you are looking at can be changed!

 

I cannot, I promise! There is no way for me to change that value on the main screen. I added the picture in my post above.

 

How about take your mouse, click on the 128 number and edit it. Use the Delete key on your keyboard and type a new number in there.

In the TPL window, the 7700HQ has a maximum multiplier of 38 so I would set the Speed Shift Max value to 38. I think the minimum multiplier is 8 so set the Speed Shift Min value to 8. That is how these values are supposed to be set by default.

 

Huh, I guess it was that obvious then. Thank you. Please save me my dignity and delete these posts!

 

You are not the only one that has had this problem so I have to take some blame for bad design. Sometimes I over simplify things.

Too simple can be just as confusing as too complicated.

 

How do you disable memory integrity check? I'm logged in as the administrator but I still can't turn it off

http://prntscr.com/nekcm0

 

I can confirm that disabling the memory integrity check is the only thing that allows you to use throttlestop on the latest version of windows. To disable the memory integrity check, you have to first do the following:

setting the following key to 0:

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

Then you can go into the windows security center and under device security, click on core isolation details and you can now disable the memory integrity check

 

Hi,
I noticed a small annoying thing in the interface.
I'm using 150% windows scale and can't access to ok button in the fivr. I have to go down to 125%.

I'll publish soon my test trial, errors, screen shots results.

 

Use 135-140%. I feel the latest update broke the display scaling on FHD panels.

 

I posted this in the GS65 lounge to see if they know of any BIOS options to change, wondering if there's something I'm missing in Throttlestop.

No matter what I set for static voltage in TS, my 8750H adds 0.0850 to it under load; goes back down on idle. Thoughts?

 

Left mouse click on any open area in the FIVR window, hold the mouse button down and you can drag the window around to wherever you want. The entire ThrottleStop GUI is like a giant title bar. I have always loved this hidden feature. Can you drag the FIVR window and access the OK button with that trick? The Visual C++ MFC libraries that ThrottleStop uses have their origins from the Windows 3.1 / Windows 95 era. Huge displays and scaling were only a dream back then so ThrottleStop having some scaling issues is no surprise.

You can use the FIVR Static voltage adjustment in ThrottleStop to request a fixed voltage but there is no guarantee that the CPU you are using will honor that request. I tried this on an 8550U and the static voltage request was ignored. Very few people in recent memory in this forum are using Static voltage. Lots of people with the 8750H are having great success by using the TS Adaptive voltage setting combined with a negative offset voltage of about -125 mV. What sort of problem are you trying to fix by using Static voltage?

 

You have to choose apply voltage immediately, otherwise it will be applied on startup if you created a task to run at start with full admin rights. You need to disable Hibernate completely, otherwise TS won't apply its settings unless you restart your PC.

 

Stability. Using offset gives no way of testing whether any of the bins on any of the cores could be spitting errors at the reduced voltage. I get that for gamers nothing crashing after basic testing means they're fine, but I run heavy financial stuff on mine. I can keep Prime running and adjust the voltage to test, but that still means that it will have to be set 0.1v higher than needed in the end. At 3GHz on all cores, there's a 8 degree difference in heat between 0.8200v and 0.9050v. However, both of those are still better than the max 2.8GHz I get at stock.

I get this isn't a fault of the program and the wonderful work you've done, I'm just annoyed with Intel's decisions. This experience is making hold off upgrading to a 9900K as well, if it behaves the same.

 

You have MSI. Look after bro @Falkentyne posts in MSI threads about AC/DC Loadline. What you see is MSI's firmware engineers screw up.

 

Hi guys, would really appreciate a hand here.
I've been trying to figure out this for the past few weeks but I haven't gotten anywhere.
I've got a Dell 7567.
It's got a 7300HQ and a 1050ti.

The issue is that the clocks of the CPU never goes below 16, even when left completely alone with nothing running. Even on a fresh install of windows!
I've tried numerous settings on TS but it just doesn't seem to work (could it be that this is a bios locked setting?).
I will be leaving my TS settings here:

 

CPU clocks varies with SST value, 150-190 cause CPU clocks to use multipliers between 8-21x. So it better to choose 128 for battery to get on-demand turbo boost to be working.

 

If you are trying to minimize power use, you are there. Note that the power consumption is very low, and %C0 in both of your examples is quite low. What you need to know is that the left column speed registers when the CPU is doing something - but in both of your examples, it is spending most of its time in resting state. You may be better to allow turbo to be on - you will get better responsiveness, and probably still low power consumption. UncleWebb has pointed out that it is actually more efficient to allow CPU to run fast, get its work done and go back to resting state. On my 6700HQ, if I set EPP=0, FID fluctuates around 32-34, but CO% remains very low, as does power consumption. Try playing with it to see what yours does. Vasudev is right that EPP of 128 corresponds to balanced mode - people often use that for battery use.
Best,

Joe

 

Hi, it seems that the minimum CPU clock (8x) is not usable by the laptop during iddle usage. The only time I've seen the laptop use clocks below 16x, is when it is power limit throttling, which doesn't happen anymore thanks to TS.

Thanks, and yes, my objective is to minimize power use.
Basically it is pointless trying to go around this lock to achieve a multiplier below 16x when the laptop is idle?

 

That is correct - the multiplier you are seeing at idle reflects brief times when the processor wakes up to run some background task (it's impossible to eliminate all of them, though you can certainly try to eliminate anything you don't need). There are long chains of messages that can help you figure out which services can be killed without breaking something. There is also a thread that will tell you how to park some cores when not in use. That doesn't help as much as you might think because the resting states (C7...) are pretty close to parked anyway.

J

 

Hi there UncleWeb

I last posted here using Throttlestop on a win7 64 bit laptop. Today I have the same laptop with windows 8.1 Pro 64bit. Here's the thing - on the win7 setup I could never get the computer properties cpu oc reading right against the Throttlestop display. Now I have the same issue but not quite.

This time computer properties shows the cpu at 2.66GHZ. The Throttlestop tray icon also shows 2.663GHZ and Throttlestop display shows the attached screenshot. Have I somehow accidently ocd the FSB? There's a 2.663GHZ in there but how?



If the FSB is ocd as Throttlestop says then I didn't know that was possible with Throttlestop which is why I bought the T7600(G) with the unlocked multiplier. The cpu runs great at stock voltage with 100% no BSODs so I'm happy but I don't know why .

 

@Mindfull - On some motherboards, changing the CPU multiplier in Windows can screw up the accuracy of some of the internal timers that Windows uses.

WinTimerTester
https://drive.google.com/open?id=1CzSrMXpFjyqr_4KBOJYb52eqi_pnlFG4



To reset your CPU multiplier, push the Reset button on ThrottleStop (to the right of the UNLK button), then run WinTimerTester for about a minute and see if the two Windows counters are in sync. They should be.

Go back to ThrottleStop, click on the UNLK button and increase the maximum multiplier by 1. Close and restart WinTimerTester. Are the two counters still in sync? They should be but they might not be. Post some screenshots if you think there is a problem.

To change Windows so it calibrates its internal timer based on a fixed frequency counter, open up a command window with Admin privileges and type in this command and then reboot.

bcdedit /set useplatformclock true

Do some more testing to see if it makes any difference.

To go back to the original timer that Windows was using type in this command.

bcdedit /deletevalue useplatformclock

Reboot and things will be back to normal.

When testing, push the FSB button in ThrottleStop a few times to see if it is consistent

If changing the FSB or multiplier screws up your Windows timer accuracy, switching Windows to use a fixed frequency counter might fix the problem. Some versions of Windows used a fixed frequency timer while other versions calibrated the main timer based on the FSB / BCLK speed. When Windows is setup like this, changing the FSB / BCLK / multiplier can screw up Windows timing accuracy. This was a big problem about 8 or 9 years ago.

I have never had access to a T7600G so I have never done any proper ThrottleStop testing. It might be buggy but at least it can be used to do some multiplier overclocking. When testing, run CPU-Z and see what it says. I am pretty sure that CPU-Z reports things correctly for the T7600G even when overclocking.

Edit - My best guess is that your default FSB speed is 166.443 MHz and you might have checked out the 16 multiplier when playing with ThrottleStop.

 

Hi to all.

Few weeks ago bought a laptop with an I7 8565u processor and 15w TDP.

Trying to undervolt it I can reach -110mv Core and Cache with no issues on any test, but TS Bench fails on single thread (It’s when the higher clocks are reached and temperature tops without reaching TDP throttling). Multi thread tests won't fail at all.

Neither Prime95 nor Memtest86 would give me a single error even on single thread, but TS Bench fails just a second from starting.

I just don’t understand so if someone could enlight me I would be grateful.

Thanks a lot!

 

If the ThrottleStop - TS Bench reports errors or crashes, your CPU is not 100% stable. It needs more voltage. Reduce your offset voltage and see if you can pass that test. If you can pass with more voltage, that confirms that there is nothing wrong with the TS Bench test. The 8550U I have can pass TS Bench with no errors.

The TS Bench test is an easy way to find out if the amount of voltage going to the CPU is adequate. Many popular benchmarks heavily use the AVX instructions so they are a good way to test for that kind of code. The TS Bench does the opposite. It does not use any AVX instructions. It is a good way to test for the majority of code that does not use any or uses very few AVX instructions. A stable CPU can run Prime95 or the TS Bench without any errors. Some users probably run a few threads of each at the same time for a true, full body workout.

The 8550U I tested was stable at -100mV when fully loaded. For 100% stable, I backed that off to -80mV. The 8565U uses the 46 multiplier when a single core is active. The 8550U only uses a 40 multiplier so big difference. The higher the multiplier, the higher the required voltage is going to be. Intel is really pushing their old technology to the limit. Why? Competition from AMD.

It is your decision. Maybe the TS Bench is excessive. Your computer might be 100% stable and never cause you any problems at -110 mV. If you have a BSOD or some other weird problem a week or a month from now, I guess you should have listened to what the TS Bench is trying to tell you. Not 100% stable.

Here is a guy on TechPowerUp that decided to listen to the TS Bench results.
https://www.techpowerup.com/forums/threads/8700k-realbench-vs-throttlestop-bench-1024m.254890/

I guess the voltage he thought was "game stable" really wasn't as stable as he thought.

 

Hey! Thanks a lot for your detailed reply.
It's more clear to me now the different type of tests being performed by TS Bench.
After increasing a little bit the voltage I can pass 1 thread test with -95mv.
Anyway I come from a ten year old laptop that guess didn't have this kind of TDP throttling so all this thing is quite new to me.
I mean, it's somehow counter intuitive that the laptop can pass an 8 thread test but fail on a 1 thread, but once you get that with 1 thread you are pushing a core to its limits without reaching the 15W TDP limit t makes more sense.
Guess I'm gonna change that according yo your suggestion.
Thanks again!

 

@pantezuma - For the next lesson, the 15W TDP limit is only an Intel recommendation. Some manufacturers like Lenovo have decided to leave these limits unlocked. With the help of ThrottleStop, you can go way, WAY, WAY beyond 15W. I just wish Lenovo would have included a more robust heatsink and fan but no complaints. The 4 core 8th Gen U CPUs are beasts!

 

Thanks for your reply.
I happen to have a Lenovo and it seems the TDP limit is unlocked.
I'm tempted now to modify it and see how far can it go...
Gonna think about it cause I wouldn't like to burn a two week old machine!
Thanks again!