The ThrottleStop Guide

@FrozenLord - On the Bay Trail CPUs, the voltage and multiplier are in the same register, similar to the old Core 2 Duo CPUs. Because of this, Set Multiplier needs to be checked so that ThrottleStop can send multiplier and VID data to the CPU. If this is not checked, TS will not do anything.

To fix Clock Modulation issues as well as the above, you need to click on the Turn On button in TS so it is out of monitoring mode. After that it can actively monitor and fix your VID and throttling issues. A lot of your problems are being caused by how you have TS setup.

Look at your TS 8.60 at the top right of the screen, just under the name of the CPU. Do you see where it says VID? That is the current VID voltage. That used to be displayed on the left side of TS but I thought it would make more sense over on the right hand side in the monitoring section. It seemed like a good idea at the time.

Remember to max the VID voltage before clicking on Set Multiplier for the first time. If the VID voltage is at a minimum value, the CPU will probably instantly crash with a blue screen.

@GreatD - I do not own a 7700HQ and different laptops use different throttling methods. I would start by lowering the CPU Core and CPU Cache voltage. -50 mV offset is a good place to start and maybe you can continue to lower that in steps of -10 mV to -100 mV or beyond. When you lower the votlage, do some stability testing along the way. Do some Google searching. There are guides out there including one in my sig. Post some screenshots of TS if you need some more help with it.

Your first screenshot shows that you have Disable Turbo checked. That will seriously reduce the performance of your laptop from 3.60 GHz to 1.70 GHz. No wonder it feels laggy. Any reason why you checked that? It also shows that Speed Shift is enabled. For more performance you could try setting the SST-EPP value on the main screen to zero.

 

I seem to have missed this crucial information.
After enabling Set multiplier, my voltage is finally being accepted.

Thank you very much!

 

You are welcome. That feature is definitely a little obscure. I should have put the little check box half way between the multi and VID to try to let users know that those values are connected. This combined with a fix for clock modulation throttling can make a huge difference to Bay Trail devices but very few users ever posted their results.

Bay Trail support was left a little unfinished. Intel completely changed power consumption reporting for these CPUs compared to Core i so TS low balls power consumption data. I figured that was not too important as long as you could stop the throttling and do some under volting.

 

Still finding time to do that benchmark on my HP laptop with undervolting.

I'm still trying to figure out why it throttles not because of temp but a time based TDP PL throttle

 

@unclewebb I just wanted to point out something.

When reinstalling Windows on certain laptops, it was not enough to install the Update for Visual C++ 2013 Redistributable Package.

If you are on a 64-bit machine, you need to install both the vcredist_x64.exe and vcredist_x86.exe executables.
If you are on a 32-bit machine, you need to only install the vcredist_x86.exe executable.

Just wanted to point that out, as even with all the Windows updates finished, I was getting a mfc120u.dll error.

 

A lot of people think that as long as their CPU is not over heating, it should not throttle. Intel thinks differently. Their CPUs can throttle based on power consumption or temperature. People need to understand that before paying too much for a laptop with a 15 Watt TDP CPU. That limit might be difficult or impossible to bypass.

It is up to users to do their homework. Depending on the manufacturer, after a few tweaks in TS or XTU, some low power U CPUs will run great while the exact same CPU in a different manufacturer's laptop might be severely limited. Buyer beware. There are consumer laws to prevent automotive manufacturers from doing stuff like this but for laptop manufacturers, it is open season. They can dream up whatever throttling scheme they want and they usually get away with it. Sad industry.

Thanks for the feedback. For the next release I will probably just include the mfc file within the exe. I prefer separate so you can download the latest dll version but most people like things simple so they can just run ThrottleStop without having to install anything else.

 

@unclewebb
May I ask, what is the difference between "CPU Cache Voltage" and 'Uncore offset" on a Kaby Lake (7820HK or even 7700HQ/7700K?).

When I change the uncore offset directly in the Bios, Throttlestop picks it up as "Analog I/O Voltage (with the proper offset)".
But uncore is supposed to be the "Ring/cache", right?
If that's the case, what's the difference between "Ring", and "CPU Cache Voltage" (that can be adjusted independently of CPU core voltage) in TS?

 

http://forum.notebookreview.com/threads/the-throttlestop-guide.531329/page-449#post-10236467

 

I saw that post when I did a search, but it still doesn't answer the question.
Namely, "Chipset"? Maybe it might have been for older systems, but as you know, @unclewebb does *NOT* have a skylake or kabylake system to test on.

Plus I did a test yesterday, taking @Kittys 's advice:
I set my cpu voltage down about 10mv and increased the "Uncore offset" by 50mv and did a prime95 AVX test and kept the same stability, although temps seemed to be the same (same ambients). So then I set the uncore offset by +20mv instead of 50mv, kept the same stability and got a 1C temp decrease compared to the original test (with no offset).

Now I don't know what that means, but clearly it's affecting SOMETHING on the CPU itself. And EVERY single search I've done on the internet says that 'UNCORE' on kaby lake is "Ring".

But in discussions here, "CPU Cache voltage" has been labeled as "ring" voltage, so this has to be clarified!
What exactly is CPU Cache voltage then in Throttlestop? L1 voltage? L2 voltage? L3 voltage?

And in that case, what is UNCORE affecting then?



The above answer about "Analog I/O voltage' being "chipset" doesn't fly. Or maybe I"m just "DIMWIT Falkentyne" or something. Because I *THINK* Personally analog I/o voltage is "Picking up" the uncore voltage! and just saying its Analog I/o when it's not analog I/O.

 

The FIVR Voltage section in ThrottleStop is based almost entirely on some notes that @Dufus sent me a long time ago during the Haswell era. I think Haswell and Broadwell are similar but there are a few changes with Skylake, Kaby Lake, Coffee Lake, etc. Some of the voltage labels that TS uses like Analog I/O or Digital I/O might in fact refer to a totally different voltage in the newer CPU families.

TS might give you the option to set some voltages as Static but requesting Static might be ignored if the CPU only supports Adaptive for that voltage. Same thing for the voltage range. You can use TS to request a CPU Core Static voltage of 1.9990 V. You hope the FIVR is smart enough that it will ignore this request.

In Haswell, uncore voltage became known as System Agent. Cache and Ring voltage seem to be two different names for the same thing. XTU displays this voltage as Cache so I used the same label in TS.

Without proper documentation, it is mostly a guessing game. When I am not sure about something, I try to leave things unlocked so users can do some experimenting. I have access to a 7th Gen Kaby Lake system for testing purposes but not sure if it has an Uncore offset adjustment available in the bios. I will try to do some more testing in the next couple of days.

 

Excellent and thank you. I noticed that decreasing my CPU vcore voltage by 10mv and increasing the uncore voltage by 50mv (was a dirty fast test) kept stability the same in prime95 and temps were basically identical. Then -10mv on cpu vcore and +20mv on uncore rather than +50mv seemed to maintain stability and lowered temps by like 1C overall, compared to baseline (the default vcore without changing uncore).

 

After hours and days searching the web, the only solution I've found is yours. Thank you very much @psyke83 Really. You are a beautiful person.

Is this actually still the only solution for this problem?

PS: Why is Microsoft trying to boicot the WinRing0 driver @unclewebb ? I don't know if it is true, but I've heard that that same driver is also used by some malware softwares...

 

@kameyo - I recently installed Windows 10 - Version 1803 - Build 17134.1 (RS4) just to make sure that Microsoft is not picking on ThrottleStop or the WInRing0 driver. I also just updated to 17134.48 and everything is OK. No problems in RS3 either. There seemed to only be a few fast ring builds where TS was not working without the helping hand from Open Hardware Monitor. What build are you using?

The biggest problem might be Windows Defender. Are you using that? I use Avast Antivirus which works well after you disable the majority of its ever increasing bloatware. Defender has been known to throw a fit or two over ThrottleStop without much explanation when it does do this. It will just prevent TS from starting.

Why is WinRing0 on the bad list? It is a powerful driver that in the wrong hands could allow a virus writer to take full control of a computer at the lowest hardware level and could really screw things up. WinRing0 has not been updated in 10 years and is likely considered a security risk. I have been contacted off the record and been told that I should not be using this driver. If TS was a commercial program, I could send a bag of money to Microsoft or get someone to develop a signed driver for TS. As a free program, I cannot afford to do that so it is WinRing0 for as long as Microsoft is in a good mood. I try to lay low and not offend too many big companies with my little program.

@Falkentyne - I had a look at an Asus Z170 bios and found a VCCIO voltage setting. This voltage is supposed to be a combination of what used to be the Analog I/O and Digital I/O voltages. ThrottleStop shows an Analog I/O voltage setting but nothing is displayed for Digial I/O. The bios allows fixed adjustment of VCCIO but TS only allows offset adjustment of Analog I/O. Not sure if I will be able to learn anything new but I will definitely play around with this and the System Agent voltage too during the next few days and compare that to TS.

 

I don't receive Insider updates in this PC, but I have the system up-to-date. Windows 10 Version 1803 (build 17134.48).
In my case, TS didn't stop working directly after the update. I started receiving the "Could not open WinRing0.dll" message error maybe 2 or 3 days after I installed the update.

Yes, actually I am using Windows Defender... Would adding TS or WinRing0 to the exceptions change something?
I've had installed AVG, or Avast, among others before. But since Microsoft launched MSE (Microsoft Security Essentials), later called Windows Defender, I started using it and never felt the need to change back to another one.

Well I guess there isn't a similar up-to-date driver that you could use instead. Lets hope they won't block it for real any time soon...

TS is quite small, but very effective. It punches way above is weight.

 

I rest my case. If Defender has an exception list, you can try adding TS to it and see if that makes any difference. I am not a big fan of any nanny software that blocks programs and doesn't have the guts to say exactly what it is doing. I was trying to do some TS testing last night on an older tablet that is still using Windows 8.1. I ran into some issues just like you are having and Windows Defender was to blame. Sometimes TS would start up just fine and then suddenly it would be blocked. I would have to create a new folder with a new name to try and trick Defender. Waste of time. Avast offers a free antivirus version too, without the headaches.

 

I see, I guess now I have a reason to change the antivirus hehe
Thanks for everthing @unclewebb Your work is much appreciated!

 

So, I don't have (or know for sure) 6th gen access, but I know that 7th gen and 8th gen tie vcore and cache voltage together. You cannot separate them. Raja from ASUS even commented that it is by design. If you drop cache voltage to say 1.05v flat and you put core voltage to 0.01v you'll still get 1.05v effective because the higher value is taken. I tested before with throttlestop too when I couldn't get undervolting to work and learned I needed cache to match core. Anybody with a TS profile where core is lower voltage than cache and is using 7th gen and onward (and possibly in effect for 6th gen too) is fooling you, and possibly themselves. It doesn't work. Here's that article by ASUS: https://rog.asus.com/articles/guides/the-kaby-lake-overclocking-guide/

As far as I have been able to tell, VID is consistent with any other reading tool I've ever used. But what VID doesn't tell is current consumption, which clearly is part of the equation. More voltage can mean less current consumption (amps of course) so temperature isn't completely scalable. I remember there was a point where I was helping a user with a haswell notebook on NBR where he COULD undervolt more than he settled on, but he needed to raise current limits to keep the same clockspeeds, and it ended up making his CPU hotter than simply using a weaker undervolt outright.

 

@unclewebb I think adding "ThrotleStop.exe" (both as process and as file) to Windows Defender exceptions list solves the "Could not open WinRing0.dll" problem. I might need to test it more, but it seems to work.

 

You're right, I fooled myself until unclewebb said I was wrong. I still get -260mV of undervolt with the cache but I'm also testing why setting JUST my core to -500mV gives me a blue screen but anything under that doesn't regardless of my cache voltage.

 

I'm using ThrotleStop for a few months now and it was all working perfectly, but a few days ago the gpu notification (monitoring) icon that i keep in my task bar (along with the cpu monitoring icon) is always showing a number temperature number, even when my nvidia is in idle (yeah, i know that it's not being used because i use another monitor that shows when the GPU starts to work). Until a few days ago when the GPU was idle it showed 0. Any ideia why and how to fix??

I have an Acer PRedator HElios 300 with a GTX 1060.

 

Maybe disable Malware checking of Downloaded files using a tweak. Hopefully, WinRing dll should work.

 

I think super high undervolts break something with the CPU. I saw someone else who was trying -500mV and then... when he raised core, VID/power consumption would raise, and when he set core back to 0 offset, he was overvolted by like 100mV.

So don't make such big differences. Intel's CPUs are not... very smart...

 

Good news. I had some time to play today and I found where IccMax is hiding.
Yes, adding this to TS is on the things to do list.

What other GPU monitor are you using? If too many apps are monitoring an Nvidia GPU, it is going to wake it up.

What Nvidia driver are you using?

Nothing in ThrottleStop that is Nvidia related has changed recently so I think the problem is either the other GPU software you are using or maybe it is a new Nvidia driver or perhaps a Windows 10 update. Win 10 also monitors the GPU these days so maybe it is the problem. What version of Windows 10 are you using.

TS was always nice because it was Optimus aware. It could monitor an Nvidia GPU without waking it up. I do not have the latest Win 10 installed on my laptop so I cannot do any testing. I hope MS have not screwed things up.

 

The Asus desktop board has a bios option so you can disable CPU SVID Support. This allows a user to adjust the offset voltage sliders in TS without any change to actual voltage going to the CPU.

At default voltage with zero offset, the TS Bench is showing 43.0 Watts. VID voltage is between 1.2377 V and 1.2389 V.



Now for some fun, I set both the CPU Core and CPU Cache to a negative offset of -1.000 V. Not -100 mV but I went all in at -1000 mV.

Intel CPUs do not measure power consumption. They use a magic formula to approximate power consumption and this formula is based on VID. When you trick the CPU into thinking the VID voltage is much less, the reported power consumption will also be much less.



VID is now between 0.3175 V and 0.3292 V. With the exact same TS Bench load, reported power consumption is now only 13.5 Watts. CPU Core temperature is almost identical which means that the voltage going to the CPU is identical in both tests. If only I had an 8th Gen laptop with a U CPU. Maybe I could harness a trick like this to drop reported power consumption and get around premature power limit throttling.

Edit - The other thing I learned is that you can use TS to under volt the Intel GPU but you have to under volt both the Intel GPU and what TS is calling the System Agent.

On this 7th Gen CPU, you have to under volt the CPU Core and CPU Cache equally but as others have found out, you can under volt the CPU Cache more than the CPU Core. If you under volt the Cache way more than the Core, you will be rewarded with a BSOD so these two voltages do not seem like they are 100% linked.

 

Any idea why TS doesn't read my CPU speed?

 

Can you think of any software you are running on your computer that might be blocking TS from accessing your CPU? An antivirus program or a sandbox feature within an antivirus program might cause this problem. Is your CPU a retail CPU or an ES engineering sample? CPU-Z should show you this info.

FYI, Set Multiplier does not work when Speed Shift is enabled (SST in green). To limit the max multiplier try adjusting the Speed Shift Max setting in the TPL window.

 

If only I had the money to buy you an 8th gen U cpu laptop.... I would so you could research VID tricking. lol

 

Master Unclewebb!!!

You can already do this!
Here!!! But very few Bioses that are not Premamods have this unlocked and accessible.

Set this to a value LOWER than 100 (100= x 1.0).

 

With what you've been showing with VID changes being an approximation, it might very well be possible that cache voltage is the number 1 limiting factor for VCORE and undervolting the cache too much simply breaks it, whereas VID simply shows whichever is the less aggressive undervolt... let's see...

I tried raising core instead of cache and then cache instead of core and with each raised (both 1.05v normally, raised one to 1.2v and left the other at 1.05v, then swapped the values) my temperatures raised as well, so it seems like it goes back to what I said before: intel CPUs aren't that smart.

It might be a little different on adaptive, however.

 

Retail 6700k
No idea what is causing but since the windows 10 spring update the value cannot be read.

Enviado do meu SM-N950F através do Tapatalk

 

I usually disable GPU usage in task manager in Process section. Possibly W10 querying periodically causing dGPU wake-ups!

If you're using Sandboxie or Hyper V or any hypervisor w/ bare metal OS or even Containments can give you similar result.
Firstly, you can try disabling hibernate and see if it works.

 

Can someone shed some light here? 8550u Here

Ok so let's give these scenarios.

1) CPU is under full turbo load. 30watts PL2 is lit and EDP Other is Lit (Ring). CPU is at 3.3ghz
2) CPU is under 1 core full load. 10watts, PL1 is lit and EDP Other is Lit (Ring). CPU is at 3.8Ghz on 1 core
3) CPU is under full Load base clock, 15 watts, PL1 is lit and EDP Other is Lit (Ring) CPU is at 2.4ghz.

Is there a limit reason that's associated with a certain CPU speed? Why is PL1 associated with a TDP of 10 when under 1 core load and full base case load at 15 watts? TDP Down and up must be it?

 

Might have had a breakthrough on my PL1 limit.

Setting the highlight value from 40-90 increases PL2 and it stays on PL2 forever. 90 being somewhere around 18.5 watts 40 being 6ish watts. Setting this value to what others have set it too doesn't do anything for me. Only changing the last 2 bits do. Even setting it to E8 only shows thermal throttling then finally PL1 sets in. Interesting.

 

Hello, I have a MSI GE62 7RE with a 7700HQ. TS works fine, but the problem I find is that everytime I turn on the laptop from suspension, the undervolt resets and I have to open TS again. Any idea of why this happens?

 

That's normal. TS needs to be running, otherwise the CPU gets reset after waking from suspend. It's a very lightweight program and you can just leave it in the background all the time with no impact.

 

Because the bios resets the CPU registers when it wakes up from suspend. Your Bios wasn't designed around Throttlestop. Load TS again and reapply the undervolt.

 

Is it still recommended to enable Speed Shift on the new i7 8750 CPU? Also, some guides recommend undervolting the GPU while others don’t. What is the consensus on that nowadays? Thanks

 

It is your laptop. You can set it up to run however you want it to run and ThrottleStop gives you access to a wide variety of tweaks to accomplish that.

Enable Speed Shift in the TPL window and give it a try. After that, play around with the EPP value on the main screen. If you like Speed Shift, leave it enabled. If you do not like it, reboot and do not enable it. No need to use something just because 51% of people like it.

Some laptops running Windows 10 will enable Speed Shift automatically. In this case, you do not have much of a choice. Microsoft has made that decision for you. Once enabled it cannot be disabled without rebooting and if Win 10 turns it on every time you reboot then you are stuck with having Speed Shift on whether you like it or not.

 

Is it possible that TS can keep the PL1 or PL2 flag on the i7 7700HQ processor when on idle?? I have continuous TDP throttle flag (dot) on TS and I'm under temp of 60 C degrees. Maybe I should do bios upgrade and reinstall Windows 10. Any ideas guys?

 

Thanks, I thought it was due to some TS or BIOS misconfiguration, since XTU sometimes used to keep the undervolt after waking up the laptop.
BTW, If I get stable temperatures, should I increase Turbo Time Limit to the max?

And another question, what does or which effects do the "clamp" checkbox next to the wattage do?

 

For some reason Task Scheduler runs ThrottleStop after system restarts but voltages offsets keep resetting to 0 after every shut down with “Run whether user is logged on or not” setting applied (ThrottleStop supposed to run in the background in this case). It starts after shut down normally only if I set task to “Run only when user is logged on” but I don’t need GUI launched every time; just in the background for applying offsets. Unchecking both “Task Bar” and “Notification Area Icon” in ThrottleStop is a bad fix because I’d have to edit ThrottleStop.ini in order to access GUI and ThrottleStop sits in Task Manager in this case. Would appreciate your help.

 

You can use/check-off Persist FID/VID values after exit under Options .

 

That is not right. When a CPU is mostly idle, it should not be setting the power limit throttling flag, ever. Does TDP Throttle ever go off? Post a screenshot of the ThrottleStop TPL window so I can see if there is anything unusual in there. Why is your Speed Shift EPP value set to 190? That high of an EPP value will reduce performance.

The two turbo power limits control when the CPU will reduce turbo boost and throttle. Normally, the CPU will only reduce the amount of turbo boost so worst case, the CPU will still be allowed to run at its default speed. When Clamp is checked, the CPU will continue to throttle some more if necessary and will reduce the CPU speed below the default speed. I do not recommend checking the Clamp option.

Follow the Task Scheduler guide in the second post of this thread and use the "only when user is logged on" option. Clear the Task Bar box and the Notification Area Icon box. No need to edit the .INI file. When you need to access the ThrottleStop user interface, just double click on the ThrottleStop icon. I keep ThrottleStop.exe in a folder called C:\Program Files (x86)\ThrottleStop so you can go there and start it or create a link to ThrottleStop on your desktop.

With ThrottleStop already running in the background, double clicking on this icon should immediately pop up the user interface. It is designed so it will not start a second instance of ThrottleStop if it is already running in the background. Now you can make changes and when you minimize it, TS will disappear again and will continue running in the background with no icon access just like before you opened it up. I actually put some thought into this feature so hopefully it still works. Every update of Windows 10 makes developers cringe. What did they break now????

 

If ANY of you have SKYLAKE or newer LOCKED PROCESSORS but you have access to UNLOCKED BIOS MENUS or a way to manually unlock them, IF your Bios has a "CPU VR Settings" area and in the Core I/A domain, has a setting called IMON SLOPE, setting this to a value LOWER than 100 (1.0x) will TRICK the CPU into reporting LESS POWER than it is using, and will circumvent Power Limit (PL) 1 and PL2 throttling completely. IMON slope of 50 is basically 0.5x TDP reporting.

If your bios has hidden menus that have this (you can look with AMIBCP 5.0.2, or on slightly older bioses, AMIBCP 5.0.1 if 5.0.2 crashes in random menus), but you cannot unlock this setting, it is POSSIBLE to manually program an IMON SLOPE forcibly into this register by finding the proper Hex offset for this (keep in mind there may be duplicates for System Agent, GT, etc, those will not work), and then writing the value in an EFI boot to the RU.EFI program. This is basically the same method used to disable the "Bios Lock" bios setting except instead of writing a 0 there, you would write the value you want for IMON SLOPE.. You guys are all on your own if you want to mess around with that.

To find the proper hex value offset for IMON SLOPE, you need to get your system bios dumped (using FPTW64 -d biosbackup.bin -bios is the easiest way), either using FPTW64 or opening a raw BIN or ROM BIOS file (not an exe) in UEFITOOL. Then search for something that is in the APTIO setup under search (IMON SLOPE, BIOS LOCK, RTC LOCK will all work) you should see something found under "PE32" at the bottom, double click that, then a field will appear at the top, right click that field, and choose extract as is.

Then open Universal IFR Extractor (another program), open what you extracted, then save it as a text file. If the file shows up in green, it worked. then you can open it in wordpad and look for the correct IMON SLOPE offset and go from there with booting to RU.EFI (secure boot must be disabled). I will NOT help anyone with this.

 

Thanks for your response @unclewebb. What is the recommended EPP value for a i7 7700HQ? Must it be higher at 240 or lower?? Please help further. I reset all my values and turned off TS for now.

 

I have recently bought a Dell 5290 and am currently awaiting delivery (should arrive in 2 weeks) that has got an Intel 8250u.
Can I assist you somehow in getting to the bottom of this?

 

Thanks for your response and thank you so much for such an awesome tool! I followed these guides and tried all suggestions before posting here. In order to get ThrottleStop running in the background at startup I have to check Start Minimized as well. Otherwise GUI appears on boot. However, with all 3 setting applied (Start Minimized box checked, Notification Area Icon cleared and Task Bar cleared) I have no access to GUI unless I edit .ini file, that’s what I tried to say in previous post. It’s not a big deal since I have correct settings ready, just wanted to know if there is another way. Will probably try “run as a service” method as well.

 

Try 128 for balanced or 0 for max Performance. If you read @unclewebb's post, you have already tested too high EPP value (190) who kill performance.

 

What Aptio AMI bios flasher did you use? Boot guard version or standard version?

 

I used FPTW64 -d mybiosbackup.bin -bios to create aptio capsule backup.
I used FPTW64 -f mybiosbackup.bin -bios to flash.
Not an APTIO bios flasher.
aptio bios flashers cause bricks. (if you're referring to AFUWINGUIx64 or something)

AMIBCP 5.0.2 to unlock bios menus.