The ThrottleStop Guide

@JaZoN_XD - Have you tried using the FIVR - Disable and Lock Turbo Power Limits feature? To the right of that feature, first click on the Install button. This will take you to the Mega site where you can download RwDrv.zip. Unzip that and copy the RwDrv.sys file into your ThrottleStop folder. Close and re-open the FIVR window and then check that magic box.

Now try adjusting the power limits in the TPL window. Check to see if you have any control over things.

On some laptops, that can fix the problem you are having. Not sure about the Carbon. This trick allows the similar C930 with an 8550U to run indefinitely, well beyond the 15 Watt TDP rating.

https://i.imgur.com/eo0J744.png

 

UncleWebb, thanks for the tip! I've put the file into the TS folder but when I check the box and close the FIVR window it tells me it cannot find the file specified!

 

Did you unzip the file first?

Is the Install button still visible in the FIVR window?

 

Yes the button is still visible, I did indeed unzip the file. In fact, the button was clickable even before downloading the file oddly enough. (Not sure how to attach screenshots on this forums yet)

 

If you unzipped the file you downloaded and copied RwDrv.sys into the ThrottleStop folder, when you close and open the FIVR window, you will no longer be able to see the Install button.



Look in your ThrottleStop folder. Make sure you can see that file.

 

I will try restarting my machine, I've tried restarting TS as well, but the button and the check box is still there. Is there a particular TS version that would work best with this?

 

Is it normal for it to throttle like this? It starts to throttle when the benchmark progress is at 56%. Long Turbo boost is 80 short is 100. PP0 Current Limit is 1023 while Power Limit is 0 and unchecked. Both turbo time limit and PP0 turbo time limit are at 64. Cache and Core are -137.7. IccMax values maxed.

 

I've been able to get this to work on TS 8.72! I might need to update

 

The 8750H has a TDP rating of 45 Watts so it is completely normal for this CPU to throttle at this power level. ThrottleStop is reporting exactly 45.0 Watts.

Did you follow the advice that I have been giving to JaZoN_XD? The part about checking the Disable and Lock Turbo Power Limits box. Go back a few posts and read all about it.

If this trick does not get your CPU above 45 Watts then there is nothing you can do. Some laptops with the 8750H can go way beyond 45 Watts indefinitely and some cannot.

 

Yep, first thing I did. Running latest version of TS. I have a HP Omen 15 - ce198wm laptop. I tried lower and higher Power Max values but no dice. It just won't work, too bad. But I am glad that I have visited this thread to check on the new bios fiasco for my laptop following the new security exploits. I would've screwed myself over if I updated. I won't be buying an Intel laptop, or any Intel computer again if they don't sort this thing out. Also, is it needed to have checked any boxes that say Lock in the TPL window? I tried with and without checking the lock box under Turbo Boost Power Limits but there was no change.

 

@geust123455 - I included the Lock option as a last resort. If something was messing with the power limits in the TPL window, checking the Lock box would prevent that from happening. If you have tried that and it did not solve your problem, there is no need to check Lock. You might have to reboot and delete your ThrottleStop.INI config file to try to unlock things.

HP laptops are known to use a third set of turbo power limits that ThrottleStop cannot get to. If that is set to 45 Watts, any changes in ThrottleStop will not get around this limitation.

If you enjoy under volting, definitely avoid updating your bios. During the next few months, I think quite a few users are going to be coming here and asking why ThrottleStop is no longer working. Going back to a previous bios might be a painful process.

 

@unclewebb - How can we protect ourselves from BIOS update via Windows Update?

Disabling Windows Update could be a solution, but i suppose many users would like to keep getting all the updates from windows (such as security updates and driver updates) except that microcode update that disables CPU voltage adjustments

 

I believe one possible solution is to keep a backup of mcupdate_GenuineIntel.dll (found ins System32) while you can, then if Windows pulls an update on you out of nowhere you either still have a chance to downgrade, or use the old microcode and continue doing undervolting.

Please correct me if I'm wrong.

 

I've been using throttlestop for years. Helped with a BD PROCHOT bug on an ASUS laptop i had years ago.

Most recently i've had success on removing the 15w "hidden" limit on a Dell Latitude 7490 with the i7-8650u. The published PL2 is 44w and PL1 is 28 but it would always drop to 15w and even 12w on occasion after being stressed for more than a few minutes. The Disable and Lock Turbo Power Limits feature worked beautifully, pulling a consistent 22w now bringing my cinebench r20 to 1575 multi 421 single.

Thanks unclewebb!

 

@unclewebb Thank you so much for your work.

I undervolted my new HP Pavillion 17 Gaming i7-9750H, Nvidia 1660 Ti Max-Q, 16 Gb RAM with great results.

and since you like pictures...

here are the settings for my performance profile. Temperatures and frequencies tend to be a bit better if the undervolt of the core > the undervolt of the cache. still workinprogress





nothing fancy here, working as intended and starts with windows






I can go lower than 45W TDP but anything higher in Power or Time is getting ignored
HP is being mean




Stock Voltage shows severe limitations in both Temperature and Power. CPU is barely able to get to the 4GHz target.


Long power target kicks in and the frequency is in free fall to 3,1GHz



Finished cycle and quite high i believe my old benchmark with stock settings and bios f13 had something like 120s




Same cycle with the Undervolt. No thermal or power throttling.




Again long power target is being enforced and the frequency stays around 3,5GHz. Temperatures and fan speed/noises are all in the green.




I guess this is the punishment for getting the new Bios this morning.



However the differences are only in the TS Benchmark. 3D Mark still is still showing similar scores pre and post BIOS flash same with CineBenchR15. So iam hesitant to roll back the old one. Any thoughts? =/

 

Download wanted patches as standalone and install in offline mode with Network disabled. Always have the door closed when you install patches. Always.

How to use WUB
Windows Privacy Dashboard
Set Network Connection as Metered
NBR Windows 10 Clean Installation Guide
Windows 10 Tweaks and Fixes (Index post #1)
O&O ShutUp10 - Do not use Windows 10 without it!

 

How do I disable/enable core c6 state (not package state) in Windows without entering bios? which MSR registers?

 

Hey uncleweb,

after some days of usage of the old version of ThrottleStop I tried to switch to the current version using the same stable settings and oddly enough, my described problem did not recur - neither after restarting the application nor after waking up from the hibernate mode. Very strange, but everything seems to work now as it should.

Intel XTU is not installed on my machine. While undervolting I only used CPU-Z and GPU-Z a little bit in combination with ThrottleStop. In ThrottleStop I'm using currently only the first profile.

Thanks anyway,

JackShepard

 

CA Microcode?
CA and C6 were also released for 9900k. Garbage microcodes. Causes 100 mhz loss of performance in Cinebench R15 and R20. Lower Gflops in LinX 0.9.6, 35000 sample size. (Please don't run this on your laptop unless you enjoy 100C temps). Terrible Terrible microcode.

BE/BA/B4 are all better. I went back to AE myself.

 

Yeah, tested them like you and had the same resaults.
Not even terrible, I would say those are

Or even switch back to pre spectre UCODE and ME FW like I did to be free from this garbage if you are on 7th gen or if on newer gen cpus stay with AE and one of the first me fw that contains less security patches.

 

I think you're selling yourself short (even though TS is free). TS is very useful. Despite losing undervolting control, I still use TS to successfully manage the poor thermals of my notebook with the TPL settings. My laptop would be a throttling mess without it.

The Core i7-8565U chip in my HP is paired to a single heatpipe that is inadequate for full CPU loads. Under its 51W short power max, it takes less than 10 seconds to climb to 99 degrees C and thermal throttle. I've reduced that time to just a few seconds, which suits my desire to retain short-term responsiveness while not letting the CPU hit a dangerously high temperature. I also upped the long power max from the factory 18W to 25W. The processor can maintain ~2.9-3.0GHz across all four cores at that wattage without thermal throttling, resulting in noticeably better performance.

Charles

 

Throttlestop is must have software on any laptop imo.

I'm thinking, when I upgrade from this Haswell ZBook, I'll take the top off, buy a 4930mx, slap a big desktop cooler on it and overclock the life out of it. I wonder how high it can go? 100 Watts? 200? The VRM is integrated right? So as long as the processor is properly cooled, I don't have to worry (much) about the system board?

But yeah, I'm thinking a new Throttlestop, paid, would do well. With the current feature set plus better throttling control (gracefully lower/up the multiplier, voltage, TPL depending on the temperature? Like the current alarm, but more granular), better profiles, and support for AMD. There's a market here, as 1000 pages on NBR and many many more elsewhere prove.

It's an amazing tool that's helped thousands of people, and I hope it's right at the top of your CV

On a related note, have you heard the noise a 4810mq makes when it reaches 99 degrees and starts to throttle aggressively? Sounds somewhat like metal screeching, very unsettling.

 

Crippling "older" CPUs is a good way for Intel to sell "newer" CPUs.

This strategy requires significantly less investment and effort from Intel.

 

Still up to the Notebook manufacturers <how to implement the security fixes>. They could just follow how the Desktop motheboard manufacturers add it into bios.

 

On a positive note my 9750H is able to reach the C10 state now. Before it was only able to get to C8. Dropped the idle power usage from 0,5 W to 0,3 W not quite so bad. Unfortunately i can't downgrade my Firmware (quote from HP support "working as intended." ...)

 

The current microcodes alone don't seem to block voltage control.
Tested C6 (SA00270 patched) on P870TM1 with 9900KS and 9900K and everything is still working as it should (TS 8.72):

Spoiler

Same on a MAG-15 with a 9750H manually updated with C6. Undervolt stuck just fine:

https://hwbot.org/submission/4300713_

 

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

The one we need to watch out for is SA-00289. There might not be too many bios versions in the wild yet that contain this fix.

For Windows Update, all of the microcode updates that I have found so far are from October. Those are pre-Plundervolt so they should be safe.

 

OK,

Skylake-S R0/S0 stepping CPU microcode M36506E3_000000D6 for INTEL-SA-00289 Security Advisory to address CVE-2019-11157 (7.9, High) security issue (6700K @ TS 8.72):

Spoiler

 

@Prema - Good to see that. Even though the end is near, under volting is still alive and well.

To celebrate, I have been working with a user to improve ThrottleStop for the new 10th Gen CPUs like the Core i7-10710U. It has 6 cores and a TDP of 15 Watts so it is definitely a candidate for some ThrottleStop love. Under volting works great. So does the Disable and Lock Turbo Power Limits feature which really made his CPU come alive.



Here is an interesting screenshot. The C10 Package C state really does exist and ThrottleStop has no problem detecting this.



For years I was never quite sure if maybe ThrottleStop was broken. No problems here. Just 8th Gen manufacturers that thought it would be a good idea to disable this power saving feature. I guess now they can claim improved battery run time for 10th Gen CPUs and encourage users to upgrade yet again.

 

That's what we wanted to see!
Thank you Kevin for continuing to pour your heartblood into this project.

Please let the community know if we could support you with a donation run towards a 10th-gen system to let the fire burn with added luster!

Maybe NBR themselves would even chime in...

 

What is new in ts 8.73 vs 8.72?

 

Very little. Just a few minor tweaks to better support 10th Gen CPUs. The download should include a ReadMe file so you can see what I tried to fix. Programming without access to new hardware is a bit of a guessing game. The latest update is working as intended.

 

ThrottleStop 8.74
https://drive.google.com/open?id=1rVVD_Fb3P_L1lzqW-_3DVLoxj6IrwEuO

New Features
- enabled Limit Reasons access for Comet Lake Core i7-10710U.
- added IGPU INI file option to set and lock the iGPU power limit.

Some users were having trouble with the Intel GPU power limit being set too low so I added a fix for that problem.

Open the ThrottleStop.INI configuration file and add this line.

IGPU=20

The first character is the letter I which stands for Intel GPU. That line will set the maximum power limit for the Intel GPU to 20 Watts. You can set this to whatever value you like. A setting of 0 tells ThrottleStop to ignore this setting. When ThrottleStop starts up, if it finds a value, it will automatically set the Intel GPU to this power limit and just to make sure no other software interferes with this, it also sets the Lock bit on the power limit register at the same time.

Because this locks the register, if you want to make a change to this value, you will have to either reboot or usually doing a sleep resume cycle will reset this register as long as ThrottleStop is not running. If you do a sleep resume and ThrottleStop is running, it will just lock this register again.

If you do not have this specific problem, this new feature might still be useful. You could set this to an extremely low value like 1 which would force the iGPU to continuously throttle. Why? Perhaps this trick could be used to free up a couple of watts for the rest of the CPU so it throttles less when stress or bench testing.

This feature has been tested on some 4th Gen CPUs and likely applies to many other generations from the 2nd Gen and up.

Edit - Here is an example of the new TS running on a 10th Gen i7-10710U.



Here is a 6 core CPU with a 15 Watt TDP. ThrottleStop is showing that it is throttling like a pig because the manufacturer has set a 10 Watt (PL2) short term power limit. What were they thinking? The long term turbo power limit is supposed to be set to at least the 15 Watt TDP rating of the CPU and for short periods of time, the CPU should be able to run at up to 25 Watts without any problems. Some laptops with low power U series CPUs can go as high as 40 Watts. In this laptop, the manufacturer has lowered this power limit to 10 Watts and performance has been killed.

Intel rates this CPU to be able to run at up to 4.7 GHz when one core is active. The reality is much different. When pressed hard, in this laptop, it is struggling to reach a third of that speed. There is no cure for this mess yet, other than to toss ones investment into the bin and buy something else. Disgusting.

FYI, the above computer running way below the Intel spec is a Dell XPS 13.

Here is the same laptop running at a very healthy 35 Watts just as thermal throttling kicks in.



Why the huge difference? Why allow the CPU to go up to 35 Watts one minute and then lock it down to 10 Watts the next? You do not have to be a rocket scientist to understand what this is going to do to gaming performance. Fast, slow, fast, slow in an infinite and incredibly irritating cycle. Who designs this stuff and thinks this would be a good idea?

 

Maybe the same company who was one of the the reasons you started up and created throttleStop!

I put in your own post #9219 where you explained the reason ThrottleStop saw the light...


All should read it.
https://drive.google.com/open?id=1tDRwD215OkMyx9BzTQRq1esBOXJrC4Dw
"Dell has long been a fan of using clock modulation throttling. If you are interested in the early days of Dell and clock modulation throttling, here is a 24 MB pdf download that tells all. This document is what originally motivated me to write TS"

Throttlegate: Dell covering up laptop issue [Updated]


See also my thread... How Dell cripple performance explained by...

And yeah, It's Disgusting!

 

Just joined the Windows Insider Slow - build 19041.21. Turns out it also updated my CPU microcode from 0x8E to 0xB4.

Undervolt is still working on the Core (stock is ~0.9v @ 2.4Ghz), but it seems that HwInfo won't report the iGPU undervolt. I believe it's still working, but I'll have to do further tests.

 

Hi people.
@Mr. Fox
Can you help me undervolt my laptop Asus rog strix g731gu corei7 9750h gtx 1660ti

Because if I set core offset -65-70mv so TS benchmark give me many errors. 256 and 1024 mode.
If i set -55mv so without error. Temperature when benchmark 1024 mode was 89c and cpu 3980hz

How I set more -130? Or take maximum performance from CPU and stay low temperature?

Can you show optimal settings?

Thanks advance

 

Each CPU is different, no all CPU's can do -130mV undervolt, looks like your is only stable with -55mV, there is not much magic to be done, its the silicon lottery.

Repasting might help you.

 

What @senso says is accurate. The ballpark average for 9750H is probably about -125mV (just a guess based on observations, not a scientific study,) which means many will be capable of something less.

An undervolt is also arbitrary. There is nothing magical about the undervolt setting number, only the end results matter. What is your load voltage? This can also vary by product/firmware, so -55mV on system A might be the same as -100mV on system B. (For example, the product of 500 - 100 is not the same as 450 - 100, even though you are deducting 100 in both calculations.) I think a fairly average load voltage for 9750H is around 1.125-1.175V, which will vary based on load temps. A CPU running at higher temps needs more voltage, and most turdbooks run way too hot no matter what you do. Some manufacturers do not adhere to reference firmware specifications. Some improve upon them, some make them worse, and others follow direction precisely, even to a point of fault, because they lack the ability to think on their own and don't care about the outcome.

And, 89°C is better than average for turdbook load temps. The errors mean the core or cache voltage is too low. You can try setting the core voltage lower than cache voltage and see if that helps. Leave the cache at -55mV and try more than that on core, see if the temps, benchmark scores and TS errors improve. The goal here is to achieve full performance and keep the temperatures under control. So, don't get overly caught up in the idea that there is anything special about the undervolt value. If your CPU runs right and the temps are good, the value of the undervolt being applied as a custom setting is totally irrelevant.

 

@Mr. Fox
Thank you.
I did like this and without any PL1 and pl2 throtling work..








My benchmark

 

Nice job. Temps are decent. And, your actual voltage (not the offset amount) is right about where it should be, too.

 

I'm regularly seeing Power Limit Exceeded, Ring Power Limit Exceeded, IA: Package Level RAPL/PBM PL2, PL3, IA: Max Turbo Limit and IA: Turbo Attenuation (MCT) all being hit shortly after starting up the machine. I know this isn't an HWInfo forum, but are these anything related to undervolt or overclock being too aggressive. Sorry, still learning with all this stuff.

 

I just wanted to share a ThrottleStop success story. I was recently helping a user on the TechPowerUp forum with his new 10th Gen Core i7-10710U.

https://www.msi.com/Content-creation/Prestige-15-A10X/Overview

https://ark.intel.com/content/www/u...0710u-processor-12m-cache-up-to-4-70-ghz.html

The Intel spec shows a CPU with a 15 Watt TDP rating but he was having no problem pushing this CPU up over 50 Watts.
He started hitting the wall at this point. The ThrottleStop log file was showing VRTEMP throttling. Not too surprising considering how far beyond TDP he was running. We should call this kind of performance, "beast mode". Nice to see that the 10th Gen has some potential in a thin and light package.

Code:

  DATE       TIME    MULTI   C0%   CKMOD  CHIPM   BAT_mW  TEMP  GPU     VID   POWER
2020-01-24  12:46:30   5.49   14.5  100.0  100.0    63521   34     0   0.5544    1.4
2020-01-24  12:46:31   6.68   13.8  100.0  100.0    63521   34     0   0.5547    1.6
2020-01-24  12:46:32   4.86   19.3  100.0  100.0    63521   34     0   0.5525    1.4
2020-01-24  12:46:33   7.18   11.2  100.0  100.0    63521   34     0   0.5540    1.9
2020-01-24  12:46:34   4.18   17.3  100.0  100.0    63521   34     0   0.5543    1.2
2020-01-24  12:46:35   4.23   13.0  100.0  100.0    63521   34     0   0.5546    1.1
2020-01-24  12:46:36  39.67   16.7  100.0  100.0    63521   46     0   1.0106    8.8
2020-01-24  12:46:37  39.00   19.7  100.0  100.0    63521   44     0   0.9551   10.5
2020-01-24  12:46:38  38.95   45.5  100.0  100.0    63521   60     0   0.9557   21.3
2020-01-24  12:46:39  37.91   94.4  100.0  100.0    63521   63     0   0.9742   49.0   PL2
2020-01-24  12:46:40  37.27   97.1  100.0  100.0    63521   63     0   0.9305   51.0
2020-01-24  12:46:41  37.11   97.8  100.0  100.0    63521   64     0   0.9514   50.7
2020-01-24  12:46:42  37.14   97.5  100.0  100.0    63521   64     0   0.9907   50.9
2020-01-24  12:46:43  37.21   97.1  100.0  100.0    63521   65     0   0.9502   50.8   PL2
2020-01-24  12:46:44  37.01   98.0  100.0  100.0    63521   66     0   1.0157   50.8
2020-01-24  12:46:45  37.10   97.6  100.0  100.0    63521   66     0   1.0099   50.7
2020-01-24  12:46:46  37.09   97.1  100.0  100.0    63521   69     0   0.9733   51.0   PL2
2020-01-24  12:46:47  36.98   97.6  100.0  100.0    63521   71     0   0.8201   50.8   PL2
2020-01-24  12:46:48  36.69   98.3  100.0  100.0    63521   71     0   0.9918   50.9   PL2
2020-01-24  12:46:49  36.82   97.5  100.0  100.0    63521   71     0   0.9708   50.5   PL2

 

Hello, it seems I now have my integrated voltage regulator VR faults disabled and have no way of enabling it back on my i7-4720HQ. It was changed in XTU, but I turned it back on before it asked for a restart of my p650se laptop. After the restart, integrated voltage regulator VR faults disappeared and integrated voltage regulator VR effficiency mode was left disabled. I can turn efficiency mode back on but not VR faults. Throttlestop shows VR faults as disabled with an unchecked box. Thought adding a check would enable it back, but still reports disabled.

Thought maybe it's set in the bios and did a reset to defaults, however that didn't enable it back. Tried a system restore, that didn't work. So I guess it's just stuck in a disabled state forever now, correct?

Update: Nevermind... should have did a cmos reset (removed battery) before bothering anyone.

 

It'll be jnterestiin what performance a 9750H or 10750H(?) Will get in a 15W package.
On another note, was that CPU undervolted? My older 8250U at -140mV still guzzled 40W under Prime85

Anyways I'm getting the i5-10210U in a few days. Am hoping for unlocked FED159A0 power limits and capable VRMs. Not expecting desktop class performance from the (presumable) single heatpipe, but I'll see how far I can take it. The Thinkpad E14 series is new and I'm not sure if they've managed to lock down enough things. Anything you want to see with TS and Comet Lake?

 

Definitely! With 6 cores hitting 50W, you need all the help you can get.

Thanks for your offer to help. ThrottleStop appears to be working wonderful with 10th Gen Comet Lake. Post lots of pics so other users realize this. XTU is probably not going to be supporting the 10th Gen U series so ThrottleStop might be the only option for users.

 

TS is working great on my MSI Prestige 14 with the i7-10710U and my Dell XPS 7390 2-in-1 with the i7-1065g7. Just trying to figure out which of these laptops I'll keep though...

 

I've heard the same thing as well, that XTU passes on supporting U CPU's, it started back with 8th gen, and here's a work around to install XTU on unsupported platforms - useful if XTU stops working for you. IDK if it will still work with current versions of XTU, but it's worth a try:

Workaround: XTU – attempted to install on an unsupported platform
Huawei Matebook X Pro - i7-8550U
https://kallelilja.com/2018/12/workaround-xtu-attempted-to-install-on-an-unsupported-platform/

"...The issue at hand however is that the newer, as of writing, version 6.4.1.25 of Intel XTU is not supporting my laptop anymore, throwing error ‘ 0x80070643‘, ‘ attempted to install on an unsupported platform‘ during the setup.

Workaround
Run the .exe and extract the .msi installer from your %temp% folder.
Run the .msi while passing the DISABLEPLATFORMCHECK=1 property.

Code:

## Run .exe installer
XTUSetup.exe
## Proceed until error message on screen - Don't exit
# Copy the .msi version of the installer package from %Programdata%\Package Cache\ location
copy C:\ProgramData\Package Cache\{275588D7-6C9D-4FB0-BBAE-2FA3F7C2DADB}v6.4.1.25\XTUInstaller.msi C:\Users\Username\Downloads\XTUInstaller.msi
## Exit the .exe installer
# Run the .msi version of the installer setting the DISABLEPLATFORMCHECK property to '1'
msiexec /i XTUInstaller.msi DISABLEPLATFORMCHECK=1

Seems simple enough, the author goes through great length's to show his work - how he arrived at the solution...just in case newer versions of XTU require changes to the work-around. Those details are at the article link above.

 

Looking forward to the post where you let us know which and why.

 

I just updated my HP Spectre KBL-G i7-8705G with a BIOS update from 0xc6 to 0xca which is the latest one according to Intel Github page. Undervolting still works but in the other thread http://forum.notebookreview.com/thr...n-e-2186m-after-firmware-update.831546/page-4 the elitebook 840 G6 with WHL-U and the latest microcode can't undervolt anymore. Is it possible that OEM has some level of control and HP decides to block undervolting only for enterprise/business models?

 

I got a new firmware update for my Surface Pro 7 i5 and undervolting isn't working anymore