The ThrottleStop Guide

Not entirely sure if this really does what you want to accomplish, but if it's a parameter in the .ini then you could use a small script to modify its content prior to starting TS. Make two scripts, create a shortcut to both and you can choose to start it with the specific parameters you want.

Use this for a .cmd in the ThrottleStop folder and it will keep the current .ini values except changing the DCExitTime value to 0 (disabling this function):

Code:

echo on

set replacestring=DCExitTime
set replacevalue=0

type ThrottleStop.ini | findstr /B /I /R /V /C:"%replacestring%=.*" > ThrottleStop_temp.ini
echo %replacestring%^=^%replacevalue%>>ThrottleStop_temp.ini 2>&1
copy /A /Y ThrottleStop_temp.ini ThrottleStop.ini
del /F /Q ThrottleStop_temp.ini
start /B ThrottleStop.exe

And a second .cmd to start TS with a value for the DCExitTime:

Code:

echo on

set replacestring=DCExitTime
set replacevalue=123

type ThrottleStop.ini | findstr /B /I /R /V /C:"%replacestring%=.*" > ThrottleStop_temp.ini
echo %replacestring%^=^%replacevalue%>>ThrottleStop_temp.ini 2>&1
copy /A /Y ThrottleStop_temp.ini ThrottleStop.ini
del /F /Q ThrottleStop_temp.ini
start /B ThrottleStop.exe

Of course, you can set the ' replacestring' and ' replacevalue' parameters with anything you like (say 'CPUColor' and '0x12AA03', respectively). Or modify it a bit to change more initial startup values at once.

The command window that's popping up can be hidden through a small .vbs, which then becomes the effective startup script:

Code:

Set WshShell = CreateObject("WScript.Shell")
WshShell.Run chr(34) & "ThrottleStop_notimeout.cmd" & Chr(34), 0
Set WshShell = Nothing

Code:

Set WshShell = CreateObject("WScript.Shell")
WshShell.Run chr(34) & "ThrottleStop_timeout.cmd" & Chr(34), 0
Set WshShell = Nothing

The resulting Options page when starting the notimeout vbs vs. the timeout vbs:

Spoiler: ThrottleStop Options tab

To make things work nicely with Task Scheduler you can also set a few registry settings a bit more lenient. These helped me to get a more consistent behaviour when auto-starting the .vbs:

Code:

Windows Registry Editor Version 5.00

[HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\Lsa]
"LimitBlankPasswordUse"=dword:00000000
[HKEY_LOCAL_MACHINE\SOFTWARE\Microsoft\Windows\CurrentVersion\Policies\System]
"EnableLinkedConnections"=dword:00000001

Made an archive with all of the startup scripts:
ThrottleStop - startup scripts.7z

 

Ok in my msi gp63 leopard with an i7 8750h I have the cache and core set to -.1582 and gpu, unslice and system agent set to -.0703.


Until 24 hours ago everything worked fine. I have every setting known to people turned to max performance and power. Prior to 24 hours ago my FID wouldn't go under 38 for any core and would even jump to just shy of 41 if allowed. Since then though the fid has dropped as low as 32 for cores and intermittently drops to as low as 24 for a few seconds.


I used to get between 1205 to 1230 in cinebench r15 but now it's between 1140 and 1190.


Did I accidentally damage my processor? Over usage? Registry file I need to change? What?


Thank you very much.

 

Ok my bad. I did read the post, but missed the red line... idiot.
Yes, launch with the Task Scheduler. The problem was the gray bax into the conditions "stop if the computer switches to batetry power"

 

Nice the Red line text... works None is idiot. Only somewhat blind
http://forum.notebookreview.com/threads/the-throttlestop-guide.531329/page-986#post-10878475

 

LMAO

I read so many things as I am trying to lean how to undervolt correctly. Maybe someone could help, as I'm hitting my limits.

I got a 7700hq and the stable setting I found are : CPU core and cache -105v, IGPU IGPU unslice and system agent -60v
1ST goal is to lower cpu temp one AC profile.
2ND goal is battery on DC profile.

So considering the very different things I read, my questions are :
- Should I try to undervolt less cache to undervolt more core?
- If I undervolt less IGPU IGPU unslice and system agent, may it help to undervolt more core?
- What is the most important to lower CPU temp?
- What is the most important to lower power consumption
I'm a bit confused with everything I fonud here and there.
Thanks

 

1 - Try both ways + as well Core and Cashe equal. Only testing (trial/error) will confirm what's works for you.
2 - Nope, I don't think so.
3 - Voltage and proper done thermal paste/pads. Be sure heatsink is in good shape.
4 - See point 3

 

Thanks
3 - of course, but what voltage setting exactly? CPU core helps more that CPU cache to thermal? Cause if it's the case, it may be interesting to try different voltages. But if CPU cache gives you equel thermal gain as CPU core, the it may be useless.
Same principal to battery gain. Does CPU core use more watt than CPU cache?

 

Normally is both needed for best temp. You can of course lower cashe ratio (Cashe clock speed) for better temps. But this means you somewhat cripple the performance (Not much but still below specs). Only trial and error will confirm best results. Remember core and Cashe voltage rail is linked. Regarding wattage... Look at Cpu Package Power. This what's counts. You find it in TS main windows/Hwinfo64.

 

So my settings are too low as well?

 

hahaha. When did you start working at the JokeBook dept. of Dell?

 

He, Fox and Johnkss are Head of Dept. for no Jokebook for life Overclocking University.

 

I have 3 questions if possible.
1) if I have my main profile set to 1°, and I want to switch to 3° which is for gaming, how do I do it manually? Is it possible to do it automatically?

2) I have seen that many change the voltage of the iGPU, should it be done for better temperature management without decreasing performance?

3) if I insert an auto-switch from profiles 1° to 2° when I reach 82 ° in the CPU, and passing to the 2° profile I have very reduced statistics (turbo boost disabled, reduced core frequency, increased undervolt to the CPU), and this happens frequently during a game, can I cause damage?

Thank you!

 

Here is a hint.



You can also create some keyboard shortcuts if you want to switch profiles using a Hotkey. Look in the Options window for the Hotkeys button to set this up.

What many people do is not important. When gaming, are you using the iGPU? If you have an Nvidia GPU and your game is using that then changing the iGPU voltage is probably not going to accomplish very much. If you only have an iGPU then you can try undervolting the iGPU. On newer CPUs, make sure you also undervolt the iGPU Unslice equally. Do some testing on your laptop and see what works best for you.

I would not waste my time with using ThrottleStop to switch profiles based on temperature. I added this feature to keep some people happy but there is rarely a legitimate reason to use this feature. ThrottleStop does not sample your core temperature frequently enough so you will end up with sudden changes in performance, back and forth.

Why did you choose this temperature? Intel rates the majority of their CPUs to be able to run 100% reliably right up to 100°C. I prefer to just let the CPU manage itself. If it is too hot, it will automatically throttle and slow down just enough so there is not a huge drop in performance. Intel thermal throttling at the hardware level is a hundred times better than trying to use software like ThrottleStop to control the CPU temperature. Let your CPU run wild. It will thermal throttle if it needs to regardless of how you have ThrottleStop setup.

 

thanks for the answers!

I will try to use the hotkeys.

I have a cou i7 8750h and gpu rtx 2070mq, so your advice is to make the iGPU undervolting as much as the cpu? I currently have stable results with the -130mv CPU, while the iGPU for now I tried it steadily at -50mv, do I still have to go down it?

I set 82 degrees because I was advised not to go over 85 to avoid damaging the components and reducing the life of the laptop.



Inviato dal mio iPad utilizzando Tapatalk

 

hey guys,

I have GS65 8RF and I was shocked how SpeedStep works on this machine.

I tried to everything except BIOS tweaks. My problem is that it does not drop frequency to low one on IDLE. So when I open CPU-Z or other monitor and my PC is in idle (Just Skype in background) my frequency and voltage goes in circle: from lowest to highest then circle repeats. I don't see any sense in 0.7 -> 0.9 -> 1.1 -> 1.25 jumps every second without load. I tried every step selection in balanced profile, made new performance profile, tried all Dragon Center options and also deleted it and checked power profiles again. I never had any issues on Desktop PC's.

The only workaround was to disable SpeedStep and enable SpeedShift with 128 value. I know it may interfere with Windows settings for EPP but it's ok for now. I wonder if I can fix it somehow with 'C State' window. In it it has most time in C8 and C2. Package C state limit has C8 set in dropdown.

I also wonder what can I tweak in TPL window.

I have changed stuff in it but didn't notice any difference yet.
I have Short disabled, Clamps disabled, Long enabled and set to 96, Turbo time limit is set to 96.

I also have SpeedShift checkbox and text grayed out in TPL does it mean that it will be enabled even if disabled in main window?

Thanks.

 

I intend to resume my tries to eventually help some people, or maybe get some advice.
I got a Gigabyte laptop Aero 15W v7 CPU 7700HQ, IGPU intel HD 630, dedicated geforce GTX 1060.
All settings I found are safe to me. I think I'll try to push later.

Stable voltages are IGPU -60mv, IGPU unslice -60mv, system agent -60mv.

2 goals.
On batteries, smooth desktop apps and long battery life.
On AC power, max speed with acceptable heat.
From what I discovered, I can undervolt CPU more on battery mode, as I use less the CPU. I suppose parked core requires fewer volts

Battery :
Park control 25% to park cores
Throttlestop Speed shift 200, disabled turbo, CPU core -125mv, CPU cache -125mv

With those voltages, I get WHEA alert while gaming.

AC power :
Park control 100%
Throttlestop Speed shift 80, enabled turbo, CPU core -100mv, CPU cache -100mv
CPU goes around 90°C when playing.
I also undervolt dedicated GPU -150mv with MSI afterburner, and got 80°C.

When I deactivate my antivirus, gigabyte control center and wifi, I go about 80% C7 package c state percent. I tried with Microsoft, gigabyte and Intel AHCI drivers. Same results. While C0% is less than 1%, I can't go to 90% C7 and don’t know why...

 

Unclewebb reporting back after taking the 17r2 apart again repasting and re padding . still the iGPU was power throttling the CPU down to 2.4ghz and holding it there .

So I thought how about upping the HD4600 multiplier up to x14 and upping the voltage to the hd4600 as well .

With amazing results I have been playing PUBG and BF4 all weekend so far with absolutely ZERO throttling what so ever the CPU now runs max 75c and doesn't downclock from 3.4ghz

Have you come across this type of thing before ? why would this solve my throttling issues as its bamboozled me LOL

MASSIVE MASSIVE THANK YOU AGAIN


BTW I have only XTU so I could change the iGPU multiplier

 

Do you guys know why in prime95 or TSBench my CPU goes to 3.4mhz only when intel says Intel® Core™ i7-7700HQ Processor (6M Cache, up to 3.80 GHz).
Turbo enables, Speed shift 80, 100% maximal state in power plan...
If I set Speed shift to 0, I see the processor going to 3.7, but still 3.4 in benchs
Thanks

 

The spec for your processor is, I believe, 3.8 for one core active, 3.6 for two, and 3.4 for 3 or four cores active. What you are seeing seems normal.
Best,

Joe

 

Did I say that. I don't think so.

On most recent CPUs, you need to adjust both the iGPU voltage and the iGPU Unslice voltage equally. If you do not do this, your voltage request will be ignored. I would leave both of these set to zero offset. When you have an Nvidia GPU, the Intel iGPU is not being used for gaming so it is not that important.

CPU-Z and many other monitoring programs do a real lousy job at accurately reporting what speed a CPU is running at when it is idle. Do not use them for this purpose.

One should not be surprised that Speed Shift is a much improved version of SpeedStep. When you give engineers a couple of decades to think about something, they usually improve things and make things better.

If I was using Windows 10 on a modern CPU, I would let Windows control the Speed Shift EPP value. That means you do not need to check the Speed Shift EPP value in ThrottleStop. On the Lenovo laptop that I have been playing with, moving the Windows power slider all the way to the right to "Best Performance" sets the CPU to an EPP value of 0. That is perfect for me. Some manufacturers might use different default EPP values. These settings are usually hidden in Power Options. Open up a command window with Admin privileges and type in this:

powercfg -qh

That will show all of the hidden EPP settings that Windows is using. If you can see them, you should be able to change the Windows EPP settings to whatever you like.

For Package C7, somewhere around 80% to 85% on a lean system is typical. For Core C7, I like to see 99% when idle but this is difficult to achieve. Too many popular programs are written by people with no regard or understanding of what their programs are doing to C state activity. Less time in the deep C states directly translates to less battery run time and more heat. The Adobe Creative Cloud developers all need a good kick. Trying to track down a poorly written driver is not easy to do.

I have not heard of your problem before. I am glad to hear that you were able to get your CPU up and running at its full Intel rated speed.

A CPU running at only 3.4 MHz would be very bad but a 7700HQ running at 3.4 GHz when its 4 cores are all active is completely normal. Head to CPU World for the official specs.

http://www.cpu-world.com/CPUs/Core_i7/Intel-Core i7 i7-7700HQ.html

 

Yes but it does not change the fact it jumps like crazy, same stuff in TS it's even more scary with 'more data' option. Still does not solve the question the FK is going on? If I didn't have SpeedShift option then I would not know what to do. Any tips about TPL part of my question?

 

At idle the numbers jump around because the processor is spending most of its time in resting states, like C7 (or parked), where clock speed is 0, or nearly so. Then a task comes along and a core or two kicks up to speed, completes it and goes back to sleep. Looking at the C0 % will give you an idea of how much time it is spending at speed - when idling, this % may be low, but will vary all over the map, depending on what is happening. You can look with task manager and see what tasks are running at idle - one way to increase battery life is to find some of these that you don't need and stop them.

Bottom line - this fluctuation at idle is normal. The only time the numbers will be absolutely constant is if you have loaded all cores to the max.

Best,

joe

 

@unclewebb Thanks a lot. I do learn much!
So if I understand well, on windows 10, beside Speed Shift, I should untrick SpeedStep in ThrottleStop too ?

About C7 Package %, I tryed with no app lauch, every non windows task killed, no wifi... That's why I thought the problem comes from driver.

 

That is the whole point of Speed Shift. It allows CPUs to adjust their MHz way quicker. Forcing a CPU to run at 800 MHz has always been a bad idea. Here is a good paper that questions the old school, slow CPU wisdom.

https://www2.eecs.berkeley.edu/Pubs/TechRpts/2009/EECS-2009-140.pdf

A CPU should run fast when it has something to do. If you want your CPU to run slower when idle, get rid of useless background tasks so it has less to do. Here is how a lean system looks when idle.



On a CPU with 8 fast threads available, the CPU has to spend very little time in the C0 state processing Windows 10 background tasks. It is all the other junk that people like to install that screws things up.

Stop being concerned about the MHz jumping up and down. It is normal. Concentrate on idle C state activity. That is what is important. As the picture shows, a fast CPU can run very cool if it has nothing to do.

As for your TPL question, I am not sure what you are trying to accomplish. Why did you disable the Short power limit? Did it solve a problem you were having? Only make adjustments if it helps solve a problem.

@michel40 - When Speed Shift is enabled, I do not think that having SpeedStep enabled or disabled makes any difference. Do a search of this forum. Someone not long ago posted a test that disagreed with that thinking.

 

Oki!
One last question tho.
With Speed Shift disabled, on AC power, high performance profile, CPU don't idle at all. It stays at max frequency, while power plan says minimal processor state 50% and my C0% is under 1...

 

That's a good thing. Didn't you read the Berkeley paper I posted. Slow CPUs are bad. Fast CPUs are good.

With Speed Shift disabled, the traditional Windows High Performance power profile had the minimum processor state set to 100%. The Balanced profile had the minimum processor state set to 5%. The 5% setting does not mean a 4000 MHz CPU will run at only 200 MHz. The 5% just means the CPU will run at its lowest speed when idle. A setting of 99% will disable turbo boost. On an 8750H, this setting will leave the CPU running at just over 50% of its rated speed. In other words, the Windows percentage numbers are not equal to an equal percentage change in MHz.

I am not sure what a Windows setting of 50% is supposed to do. It doesn't matter. For a modern CPU, enable Speed Shift and adjust the EPP variable to suit your needs. I like setting EPP to 0 but a setting of 80 makes for a good balanced profile.

I am guessing that the engineers at Microsoft did their homework and read that Berkeley paper I posted. With or without Speed Shift enabled, it should take very little load to get a CPU up to full speed. If Windows 10 is more sensitive to load, that is a good thing. The old days of locking an idle CPU to 800 MHz has proven to be a myth that needs to be forgotten. Modern CPUs and modern operating systems go out of their way to avoid this. Running a CPU at a speed where it is inefficient wastes power and should be avoided. That is why Speed Shift was invented. When a CPU has something to do, the sooner it gets up to full speed, the better. That allows the CPU to spend a greater percentage of time in one of the low power C states like C7 where power consumption is minimized.

 

HOLY MOLY!!!

This post is going to be a combination of how much throttlestop and repasting helps, but a little bit of explanation and help needed!

So, background:

Before Throttlestop, I would hit 97-98 degrees celsius running the torture test on Prime95 for long periods of time, and severe throttle would take place.

After Throttlestop, I would hit about 94-95 degrees celsius, still some throttle but not bad.

What I did:

I just repasted my XPS 9570, the factory thermal job was HORRENDOUS! I used MX-4, and did the pea method for the CPU, and the line method for the GPU. I also added three layers of thermal pads to the board above the CPU and GPU, and three layers on the heat sink right next to the fans. I wanted to touch as much of the bottom aluminum panel as possible to use the aluminum as a heatsink.

The results?!:

A combination of very, very good, and odd. Running Prime95 for twenty minutes, after repasting and undervolting with Throttlestop, I now hit only between 80 - 85 degrees celsius. That's amazing. The only odd thing in all of this, is it seems like the FID is limited to 3.2, and the processor is only running at about 3ghz! Why is that? It seems to even be limited to that speed when I disable throttlestop too, so it can't be that!


Attached are how the thermal pads are laid out, and my throttlestop settings. Tell me what you think. Thanks guys!

 

For maximum performance, the Speed Shift EPP value should be set to 0. You have this set to 128. It can be adjusted on the main screen of ThrottleStop.

If you want a slower CPU when lightly loaded / idle, use an EPP setting of 80. This should let your CPU get up to full speed as long as it is not throttling. Open up Limit Reasons if your CPU is running slower than you think it should run and find out what the problem is.

I would also check the FIVR - Disable and Lock Turbo Power Limits feature and increase your turbo multipliers now that you have your temps in control. Let's see what it can do!

Edit - An Intel GPU voltage adjustment needs to be combined with an equal iGPU Unslice voltage adjustment. If you do not do this, your request will probably be ignored.

 

Ok! Did what you told me - everything!

Raised the turbo limits to it's maximum capacity of 4.1, turned the EPP to 80, and clicked Disable and Lock Turbo Power Limits.

It's running as we speak, at a stable 3.6ghz - hovering between 85 - 90 degrees celsius!

Here is the Limit Reasons present.

Thank you so much! Your help is so appreciated by this entire community.

 

@TSE - Limit Reasons shows PL1 glowing red. That is the reason for throttling. Push on the TPL button. What is the Turbo Boost Long Power Max set to? Try increasing that to see if you can go faster and make the PL1 throttling flag go away. When PL1 goes red, it automatically triggers the RING EDP OTHER to go red too so no need to worry about EDP OTHER at the moment.

Some people have said that long term, Dell limits the 8750H to 45 Watts regardless of how ThrottleStop is setup. Never hurts to try to go further.

For the 8750H, the default Turbo Ratio Limits are 41, 41, 40, 40, 39, 39 from top to bottom in ThrottleStop. You can decrease any of these values if you need to reduce your max CPU speed to control heat. Increasing any of those limits beyond those values is pointless. The 8750H is a locked CPU and cannot be overclocked.

It is always fun helping people get a few more MHz out of their CPUs.

 

I set it to 2.50, do you think that's good?

And I think I'll limit my Turbo Clock to 3.6 - I notice if it jumps out to 4.1ghz right away it stays at 3.0 - 3.2, but if I limit it to 3.6, it stays consistently at 3.6. I think it's because when it hits 4.1ghz the heat is too much for the thermal design of this computer and the heat design can't catch up. Just a wild, yet somewhat logical guess.

 

I've been using ThrottleStop on my Dell Precision 5510 (nearly identical to the XPS 9550) for a few months with no problems. I recently bought a ThinkPad X1 Tablet (3rd Gen), and I'm trying to setup ThrottleStop on that device as well, and I've run into an issue that has never occurred on 5510, even though they're both running the exact same build of windows.

On my ThinkPad X1 Tablet, I hit install next to the option "Disable and Lock Turbo Power Limits" and added the RwDrv.sys file to the ThrottleStop folder as instructed (same as on my 5510). After checking the "Disable and Lock Turbo Power Limits" option and hitting OK, I get the error message "Error starting service: A certificate was explicitly revoked by its issuer."

I found one other instance of this error message searching this thread, @unclewebb and others suggested that it could be a change to Windows 10 that removed support for the driver or that it could have something to do with Windows defender or other security features. I don't think this applies, since my Precision 5510 does not have this issue, and ThrottleStop has correctly unlocked the CPU power units as verified by HWinfo and there are no error messages. As I said earlier both PCs are running the exact same windows build. In fact both machines just installed a cumulative update, bringing each up to same new build number. My 5510 is still working correctly and my X1 Tablet continues to have the same issue.

So, it seems pretty clear, it's not a result of a Windows feature change. Unless my 5510 is keeping some older configuration settings that are no longer user with fresh windows installs. The 5510 had a clean install of Windows 10 Pro a couple of years ago. The X1 Tablet has identical user-facing GUI settings to the 5510. The X1 has a Core i7-8550U and the Precision 5510 has Xeon E3-1505M v5 (very similar to Skylake i7). There are probably some additional security features in the 8th gen CPU, but I doubt those have much to do with certificate checks within windows, or do they? The other major difference is that my X1 Tablet has a Bitlocker encrypted SSD.

I've tried copying the ThrottleStop folder from my 5510, in addition to installing it the normal way, and the results have been the same. I've also tried adding the revoked certificate to the certificate store on my X1 Tablet. That didn't work, and my 5510 does not have that certificate in it's certificate store anyway.

Could anyone point me in the right direction? Some settings to compare between the two machines would be helpful.

Both machines are currently running OS build 17763.437
They were previously both running 17763.404

 

There was no need to change the PP0 Turbo Time Limit. Here was my suggestion.

If you are happy with performance then everything is OK. If Dell would have included an adequate heatsink and fan, your CPU would have been able to run at full speed without overheating. Relying on thermal throttling to manage the CPU speed works for Apple so I guess Dell decided to copy them.

@TechCritic - Are you using Windows Defender? It has a bad habit of randomly blocking programs or drivers that it does not like. It never informs users of what it has done. I recently installed TS 8.70 on an 8th Gen laptop with a Core i7-8550U.

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

It is running Windows 10 - 1809 with all updates and the Disable & Lock feature works great. I always avoid Windows Defender and run the free version of Avast instead. As long as you avoid the Avast bloatware that they keep adding and stick to the basics, Avast is very efficient.

 

It really is a shame! We have these super powerful laptops, but even in my case of putting some serious work into undervolting, repasting, and re-engineering it with thermal pads I'm still only running at about 90% speed!

Did laptops always throttle, and was undervolting ever an option back then? I remember vaguely as a kid coming here to this forum (I was like... 12) and people complained at how hot the Pentium 4's got, and there was no real thing such as integrated graphics to help cool the computers.

 

A few tips IMO - YMMV:

1. Your max core temps vary a lot (>10*c). Run Prime 95 for a while and you should expect real-time reported core temps to vary by only 2*c or so. This is usually from uneven paste or warped heatsink.

-- You can check the heatsink is not warped. If they are not you might very gently try to adjust them but the heatsinks are incredibly delicate so even the pro guys have broken them. Maybe RMA them if warped is safer.
​

-- We usually repaste several times. So repaste, clamp down the heatsink, then remove the heatsink to ensure we are getting even paste distribution. Try a few times. This gets core temps to be very even.

​

-- Pushing the heatsink down with foam is not very helpful over time. This is because the floppy case bottom is constantly changing pressure on the CPU & GPU so thermal compound will get squeezed out and exposure to air. Remove those thermal pads

2. Thermal pads at the vent radiators are suboptimal because hot radiators are much more efficient at transferring heat from metal to air, so cooling the radiators may not be as efficient it seems. Remove those thermal pads.

3. The bottom cover is a poor heatsink. Has limited ability to absorb heat for a few minutes. And then can not easily transfer that heat anywhere. Also a hot bottom cover will superheat the cool air pulled in by the intake fans. This kills the efficiency of the XPS' cooling system.

4. The XPS' power mosfets have virtually no factory cooling, so this is the real thermal bottleneck. Some people pad these to the case bottom. Good thermal pads will make the case bottom very hot very fast, and I found the keyboard too hot to use. Also the fans were less efficient. A few people used lousy thermal pads from power mosfets to case bottom; that provided limited cooling to the mosfets and did not overheat the case bottom. Trial and error but most people could not get the right balance here.

5. There are plenty of XPS threads on thermals for the very similar 9550, 9550, 9570, so we have hundreds of benchmarks and experiments for you to review.​

 

Hey,

Thanks for the feedback and help.

I am 95% sure I didn't warp the heatsink. It came off pretty easily and went back on pretty easily. Additionally, I screwed it back on very evenly between the screws, little by little.

I've read everywhere else to NOT lift up the heatsink after you put down thermal paste. You are the first one to say lift it up and check for even distribution. I will investigate this more.

I will remove and play around with the thermal pads and report back to this thread in the next day or two with the results - I think you are most likely right on the thermal pads sitting on the heatsink, and possibly the ones sitting on the board. I've read a few places this is good, and other places it's not. Not quite sure know who to trust.

 

I don't think you warped the heatsink. However, Dell has sold a lot of warped heatsinks. It seems Dell stepped up their game with some of the more recent 9570s. Getting individual CPU core temperatures close during stress testing is a good clue that your heatsink is flat and thermal paste application is thin, even, and with complete coverage.

Once you remove the heatsink, you need to clean the thermal paste and reapply. This is the best way to assure your repaste is spreading thin, evenly and completely. This is a popular method.

Trust no one. Test for yourself and verify. Part of the problem is wide factory variance in CPUs, heatsinks, and thermal paste application. Part of the problem is that some owners are more careful about thermal mods.

 

Oh, I see what you mean now - the individual CPU core temperatures vary a lot. This for sure isn't normal? Should I maybe just try to remove my thermal paste job and redo it and see where it goes?

I'll have some time tomorrow to work on this- both the thermal paste and thermal pad work.

On a semi-related note, my battery life has gone to hell and my fans keep spinning up and down no matter what since yesterday - I suspect when I installed Dell Power Management tool. I think I need to do a fresh install as well, because even when I uninstalled it I am having the same issues. Ugh.

Thanks for the help.

 

No. Run a CPU benchmark for a while, like Prime95. Look at real-time core temps on the main window of ThrottleStop. If all cores have similar temps when running all cores of the CPU hard for a few minutes, then your thermal paste distribution is probably even and your heatsink is probably flat.

If you see real-time core variances that are large during benchmarking, consider additional steps, like inspecting your paste job, cleaning and repasting.

Dell Command I Power Manager allows you to change fan scheme. Go to "thermal management" tab and select "optimized" or "quiet". When you uninstall, I think the settings stick but can't remember to be frank.

Also you can boost battery life with SST (ThrottleStop main window) higher. I run around 80 for everyday use but 255 might give you even better battery life.

You can bring these XPS thermal questions to the Dell forums as they are not so ThrottleStop related. http://forum.notebookreview.com/forums/dell-xps-and-studio-xps.1049/​

 

May I just say @unclewebb how much I miss and appreciate Throttlestop now its gone.

I've upgraded to Windows 1903 and winring0.dll no longer works as previously mentioned. Laptop runs hotter than ever with fans on for a significantly longer time and I've lost around an hour of battery life.

Thank you for all of your hardwork over the years - I know this is all of your control but I just wanted to wish you all the best and thank you for my past few years as a satisfied "customer"

 

That's a damn shame if true! Is there going to be a work around? Intel XTU has nothing on Throttlestop.

 

It's reported further up in this thread but this is the exact error I receive:



I have tried everything including a clean install and completely disabling UAC but it doesn't help as far as I'm aware.

Agreed - XTU is so bloaty compared to the lightweight, fast and effecient TC.

 

I'm using TS on 19H1.30 build w/o any issues. Do one thing; Copy TS's ini file somewhere and delete the folder and re-create the folder on another drive. I think Win 10 wrecked the permissions. They do it from time-time.

 

Edit: Ignore this post - my last point triggered a few cogs in my memory and I figured it out. I'm leaving this here in case some of the diagnostic information is helpful to somebody.

This is not working on Windows 10 Pro Build 18362.53 (x64).

Exact instructions I'm using:
1. Download fresh copy.
2. Using windows explorer/ 7-zip extract to a new folder on a secondary drive (the drive is not formatted with NTFS like my boot drive).
3. If using windows explorer, manually go to folder and unblock downloaded files (right-click properties, uncheck the block).
4. Run TS 8.70.6

I am not trying to restore my existing settings at this stage in order to isolate the problem.

Are you perhaps running a different flavour of Windows 10 (enterprise/ home)? Have you changed any of the windows security settings?

 

I'm replying in a new post for visibility. @unclewebb would you please sticky this to the first post for visibility.

This applies to people who are receiving this error on trying to start TS 8.70.6 on Windows 10 Pro 19H1 Builds (as of writing Building 18362.53):


I've cracked the problem. Users who who purchased their computer after the Fall Creators Update have Windows 10's Memory Integrity Enabled by default. Throttlestop obviously changes system memory registers and Memory Integrity likely blocks this function. Read more about memory integrity.

NOTE: Turning of Memory Integrity is a serious matter. Do not turn it off if you do not understand what you are doing. You may not be able to turn off Memory Integrity if your computer is joined to a corporate domain. Turning off memory integrity may violate your workplace's IT security policies.

The easiest way to disable memory integrity is via Windows Security as follows:
1. Double click windows security in the tray or run it by searching for it by name.
2. Select the Device Security Tab.
3. Select "Core Isolation Details"
4. Toggle off "Memory Integrity"
5. Follow the prompts to restart your computer.
6. Throttlestop should now run as intended - if not try the usual recovery steps: (1) move the throttlestop.ini folder somewhere else to reset settings and (2) try a fresh install.

 

@illuzn - Thanks for your detective work. I included a link to your post above in the first post in this thread. It is near the top so it will be easy to find.

Update 2:

I toggled Memory integrity on, rebooted and ThrottleStop started OK. I do not have any problems with the WinRing0 driver whether Memory integrity is on or off. This is while using Windows 10 - 1809.

Starting with Windows 10 - 19H1, it looks like the Memory integrity setting can completely block the WinRing0 driver which will prevent ThrottleStop from starting.

If Memory integrity is enabled, I do see a bug where ThrottleStop reports 0.00 MHz. I assume that TS cannot calculate this correctly without having memory access.

I also discovered that having the Memory integrity setting set to On, prevents the ThrottleStop FIVR - Disable and Lock Turbo Power Limits feature from working correctly in Windows 10 - 1809 and likely any newer Windows versions.

I get this error.

Error starting service: A certificate was explicitly revoked by its issuer.

 

Amazing work!

 

I have never been able to understand why many users want their laptops to run slow and sluggish when lightly loaded. A slow CPU might look like it is very efficient but is it?



On the left side, Speed Shift EPP is set to 84 so the CPU is running nice and slow. On the right side, Speed Shift EPP is set to 0 so the CPU can run at full speed. When idle, whether the CPU is running slow or 3 times faster; power consumption and core temperatures are almost identical. Any difference is so insignificant that it is difficult to measure accurately. A slow CPU needs to spend more time in the C0 state processing the background tasks which means less time in C7 at 0 MHz. When the C states are enabled, a slow CPU seems pointless.

 

Do you mind showing us the corresponding C state window for those 2 setups? I'm running a Microsoft Surface which is supposed to be Microsoft showing other manufacturers how laptops should be properly configured but my CPU never enters a package state lower than C2 - the individual cores will hit C7 but not the package. I'm running a very similar CPU i7-8650U so would be very interesting for me to see if it's just me.

Also, for what it's worth, I get very similar idle CPU usage and that button also shows C10 on it. Not sure what else I can do to tweak it.

 

I've noticed that "battery monitoring" option doesn't stick. Trying to deactivate it, but it stays.

For people that having a heat problem as I do (90° even with Trottle stop) while playing games, I did limit FPS (what the point of having more than 60fps on a monitor that doesn't support it) with MSI afterburner / RivaTuner, and it helps too : now I'm at 80°.

From my experience, lokk like heat is mostly generated by cpu core & cahe while gaming (I got a dedicated gpu), and I undervolted them to -100 max to me

EDIT : Ok I noticed that if I step EPP to 0, I can undervolt more without whea... Kepp trial and error, and let you know after!