The ThrottleStop Guide

@ThatOldGuy has written a guide to auto start XTU. You can get all XTU features in TS w/o impacting system performance.

 

By default, is SpeedStep On or Off ?
I have a Kaby Lake i7-7700hq, I'm pretty sure it was ON when I installed TS for the first time. Today I had to reinstall it ( deleted throttlestop.ini and restarted my laptop ) and SpeedStep was OFF when I launched TS.
I've also noticed that my clocks remain at 3.7Ghz (which is the Max) most of the time even while browsing the internet.

 

Default on, but as the 7th gen has SST, SpeedStep doesn't matter because it has been superseded by SpeedShift

 

Did you open up the TPL window in ThrottleStop and try increasing the Package Power Limits in there? If you tried this and it is not working when stress testing then at least post a picture so I can see the TPL window and the throttling that you are seeing.

Is a fast computer a bad thing? The original purpose of ThrottleStop was to make sure your CPU could run at its rated speed and with an unlocked processor, maybe a little beyond that. The important thing to watch when your CPU is idle is how much time it is spending in the various C States and what C States it is using. That data is extremely important to lowering power consumption. CPU speed, not so much.

If you want a slower computer when it is lightly loaded, I would recommend enabling Speed Shift on your 7th Gen CPU and I would set the Speed Shift - EPP value on the main screen to 80. If you do this, SpeedStep enabled or disabled does not matter. As always, if you need help, post some pictures. I am a narcissist. I need to see TS pictures plastered all over the web, especially in this forum here. It also makes problem solving a lot easier.

 

Hey unclewebb, a friend of mine is experiencing a weird issue on his Dell with an i5-8250U and a GeForce 940MX 4GB
His FPS was dropping when gaming, at seemingly random intervals. It seemed like it was hitting the PL1 during gaming and BD PROCHOT was acting up to me, so I suggested him to get TS and enable logs so I could see what was going on with his system.

At first I told him to disable BD PROCHOT and try playing. It didn't help, so he got a log for me.

https://hastebin.com/boxogodivo.css
It was clear that it was hitting the PL1 a couple of times. I also noticed that the C0% peaked a couple of times. Temps seem fine

I told him to limit his CPU to Multiplier 30 (like I did with mine, to keep a stable frequency when playing) and set SST to 0, and check how it was now

https://hastebin.com/adutexonag.css
He told me the lag became a bit more frequent and harder now. Used to dip to 25FPS, but now it goes to single digits. The logs show that the frequency is now stable and temperatures are reasonable (his PROCHOT is 97°C). The PL1 seems to want it around 12W for some reason

I told him to install Intel XTU and increase the TDP limit there. It was set to 20W, and we increased it to 28W. We also increased the Turbo time to 64s. I also got him to log the GPU temp now. He closed XTU afterwards

https://hastebin.com/qiguwegolu.css
Logs look good, but to my surprise, he told that the FPS drops became even more frequent, making the game rather unplayable, in contrast of bothersome as before my tweaks. I don't see any Limit Reasons logged and temperatures seems alright.

Then I told him to return to the default profile (maximum multiplier achievable, SST to 128) and try logging again.

https://hastebin.com/qodacocixe.css
Now he told me the game doesn't drop FPS anymore, but that's because the game seems to not get past 10FPS anymore. It's perpetually slow and completely unplayable now.
I've noticed that the GPU is lowering the temperature in the process, even turning completely off sometimes. The only thing I can think of is a hidden 80°C limit that throttles everything, even with BD PROCHOT unchecked in TS.
Other than that, I'm completely out of ideas.

What is possibly going on on his system?

 

@GTMoraes - Thanks for posting some log files. Much appreciated.

I think the biggest problem is that his 940MX does not have enough muscle to play the game he is playing at the resolution and settings that he is trying to play at. Might want to log the GPU speed next. The GPU is either throttling or perhaps simply inadequate. Maybe try GPU-Z for logging. Getting the CPU to run at its rated speed might be taking power away from the GPU causing it to slow down.

 

The game works fine, then suddenly and randomly dips to 20ish FPS, then return to what it was previously. It seemed to me to be momentarily throttling. He told me it's not related to the scenery or events in game. It's completely random and might happen even if he's standing still (He's trying to play Rocket League)

I'll try logging the GPU MHz and usage next. I didn't think that allowing for the CPU to run faster could starve the GPU resources. Seems to be the case, because at every progress I thought I made with CPU performance and limits, the game performance decreased substantially.

My first thought was some invisible limit, but if there's a "power pool", the CPU might be taking it all for himself and GPU is unable to perform correctly.
I'll try suggesting undervolt and even lower frequencies (2.4GHz perhaps?) and see if that helps somehow

----

I've now seen in HWiNFO that the GPU too have well defined limits. I'm gonna keep an eye for that

 

The situation seems to be so odd that there is something not work as our expected.

Alternatively, you can simply run GPU-Z with LinX to check the working status during high workload. As the picture in "Sensors" tab of GPU-Z, check the status of green circle is maximum or not all the time, GPU Core Clock, GPU Memory Clock with checking the corresponding temperature of the GPU and its Power used, also check the graph in purple circle can be kept as maximum steadily or not during running Render Test.

 

Very interesting stress test. I ran it on my laptop just out of curiosity and it surely did stress! It's the first time since I repasted this thing that I've heard the thermal switch kicking in and pushing the fans to 100%. Right on time too, as soon as the CPU hovered 90°C, it kicked in and helped it lower to 80°C.
Though the Render Test from GPU-Z was quite unstable (it was locking up, and when it locked up, the GPU Core clock reduced, inducing hiccups in the graph), I managed to stress the GPU through OpenCL with Luxmark. It was nice to see how much the OpenCL process impacted in performance (CPU performance fell from ~120 GFlops to 35 GFlops!)



I let it run until LinX finished (which took ~8:50s under that condition), and it has been very stable at 81°C GPU and 80~85°C CPU. However after 8 minutes I noticed a couple of new icons on Limit Reasons on ThrottleStop I've not seen before, "VR THERMAL" on both CPU and GPU. I guess it's the VRM thermal protection that kicked in. I was about to put some heatsinks on it, but given it's an absurd load, I can safely say this system might not encounter these limits again in its whole lifetime lol

Good tip on GPU-Z graphs. The MX150 has a bit more info on this Sensors tab, and it clearly shows what's throttling the GPU in my case: VRel and PWR, Reliable stable voltage and Power limit respectively. They're normal to have, so this means the GPU isn't being throttled by problems.

I'll push his system later and see what it squeaks

 

Glad to hear you can get some breakthrough in this issue. You can put some heatsinks on the MOSFETs and the coils of the VRM part to see it can improve or not. However, as my experience, the heat dissipation from the VRM is quite high, so you can see all display cards do not use the separated heatsinks and they use a large heatsink to cover all VRMs' heat components and the memory chips or use the main heatsink shared with the GPU. The main point I want to say is, the cooling method should be active way, that is there is a fan, to cool the heatsinks. Otherwise, it may not help so much.

In fact, it is not easy to do that in the laptop, although I can see some people to work it out using copper duct with small copper plane to let the heat transfer to the main fan on the laptop, not many people have this technique to do such DIY on the laptop in that limited space. Thus, if the back cover of your laptop is metal, I think it should be, using the back cover to cool the components is far easier.

Recently, I did that DIY on my laptop and it gets some help to dissipate the heat from the CPU and its heatpipe with the not heavy loading, at least, the time with turning on the CPU fan is less than before and the fan speed is lower than before too. Also, by this means, you can simply cool your back cover of laptop in order to cool the hot components inside your laptop.



The blue one is thermal pad, the thermal paste on the heatpipe is Liquid Ultra and Gelid GC Extreme.

So you can try this method if there is no any other method you can try or achieve.

 

Using the back cover as part of the heatsink is pretty ingenious. However my laptop's bottom lid is plastic.
Also, the temperatures are pretty alright in my usage scenario, and I've never seen the VR thermal limit before. I don't think I need to tinker much with it.

I'll have a look on my friends laptop today, and will report back

 

@ Vasudev i tried this https://gist.github.com/Nt-gm79sp/1f8ea2c2869b988e88b4fbc183731693 which you have posted for unhiding the advanced power options , it did unlock a lot of option but i noticed that my cpu i5-7300hq is no longer entering c7 ,c8 states which it used .
also this lead to the min pkg power to go from 0.8-1w to 2.2+ w . do i have to change some things inside the power options to fix this or is there any way to revert this ?

 

Normally C states are unaffected since you have an HQ chip which is locked.
Is hibernate turned off? you can use powercfg -h off to turn hibernate off and reboot the PC. See if it helps or not.

 

ACHI set to lowest please.


Or try the power saver option. And update your ACHI/SATA Controller + Chipset drivers + any SSD/Nvme drive controllers.

 

OK thanks.
I use powerplan switcher to toggle between Balanced and Power saver based on Power source.

 

@Che0063 @ Vasudev OK so i waited before replying so that i can properly test to find out what is happening and i have no idea .
the hibernate is off and the ACHI was already set to lowest because i unlocked those option thru the registry hack already.
Here is what i observed when i restart the system , c8 & c7 states work for a while (couple of hours i think) and after that i don't know what is happening , c3 is the highest c state my cpu is going when it's idle. I recently updated my realtek audio driver to a latest non oem one, i thought that driver was the issue and reverted to the oem driver & the problem is not fixed .Hoping someone here can help me fix this . My laptop is HP Pavilion bc-200 , i5-7300hq.

how do i add images in the post ?? image of the c states https://imgur.com/a/zQLPSRM

 

Your PC can go to C10 deeper states for better power savings. I think some process is preventing that.

 

hi everyone.
I just wanted to ask if theres a way in throttlestop to limit the max clock speed of one core?
I have a 8750H and with a -100mv undervolt the temps are great but one core (core 2) always seems to run 10 degrees higher than the others.,
Is there a way to do this or is it just down to a ****ty paste job from dell?

 

For the 8750H, all active cores are locked to the same multiplier. That means it is impossible to adjust the multiplier or voltage individually on a core by core basis.

The problem you have is not always a bad paste job. Intel uses temperature sensors that are only accurate to approximately +/-5°C. A little bit of variation from core to core is not unusual. It can be difficult to equally load the cores to come up with a fair temperature comparison across all cores. Less than perfect temperature sensors are just as likely as a poor paste job. Sometimes it is a bit of both.

If none of your CPU cores are reaching the thermal throttling temperature then there is nothing to worry about. If that one core is getting hot enough to cause throttling, it will slow down your CPU. At that point it would be a good idea to see if you can improve the paste job. Also have a look if the heatsink is square to the CPU. Some heatsinks are far from perfect. Mass produced laptops are full of flaws.

@Docktor - Completely shutdown your laptop, no hybrid shutdown, restart it, let Windows settle down and then see if any of your deeper C States start working. After that, try using stand by mode, resume and see if any of the deeper C States are working and do the same test by going into hibernate mode. Tracking this down can be very difficult after you have made a bunch of changes to your system. Very few laptops seem to be using package C9 or C10 but a few use package C8. Maybe someone with working package C8 can post all of their Windows power plan settings so you have something to compare to. Open up a command window and type in this:

powercfg -qh >C:\power.txt

-q is for query and h will show the hidden settings.
> that will redirect the output to a file called power.txt on your C: drive

Edit - Also switch to battery power when doing C State testing. Sometimes you will only get the deeper C States when running on battery power. At least your core C States are working well. 99% in C7 is fantastic.

Put html img /img tags with [ ] brackets around the direct link that imgur gives you to show the pic.

 

@unclewebb could you give me any light on how ThrottleStop reads the Package C State limit?
@Vasudev my Package C State never goes down to C9.

 

There are several counters in the Intel Core i CPUs that count upward the moment a core or the CPU package enters a C state. These counters count upward at the same rate as the default CPU speed. The Core i5-8250U in your sig has a default speed of 1.60 GHz so these C state counters will also count up at that rate. Pretty easy to convert this data into a percentage of time that the core or package has been in that C state and this data is extremely accurate as well.

Code:

Package  C8 - MSR 0x630
Package  C9 - MSR 0x631
Package C10 - MSR 0x632

You could try running RWEverything to have a look at these registers. If your CPU is idle, you should see the MSR for package C8 counting up very rapidly. For C9 and C10, it is likely that these MSRs will contain zero. If that is the case, that would mean that your CPU has not spent any time in these two states since you booted up.

The public documentation is not clear if these MSRs still apply to the newer CPUs. Based on your screenshot, it sure looks like Intel is still using the same MSRs and these MSRs have the same meaning that they used to have. Intel tries to be consistent. They do not tend to move things around to different MSR locations unless they have to. They put some effort in so as to avoid breaking working programs. The MSRs for the other C states have been the same since the original 1st Gen Core i was released in 2008.

Here is a picture that I believe is from a dual core 7th Gen 7500U. Lots of C state residency time. Nothing unusual but no C9 or C10. Those 2 are like BigFoot. Rarely seen in the wild.

 

Wait what?

 

In RWEverything, click on the User button (second button from the top left) to enter in the MSR number that I talked about. Your screenshot shows "too much information". My brain is going to hurt.

Here are some random MSRs on my User List.

 

@unclewebb here are my observations after some testing : booting from shutdown c8 state works , if i wake the system from sleep c3 is the highest c state . I have never seen c9 or c10 active c8 is the highest/deepest it goes for me which is still pretty good considering the pkg power goes as low as 0.9 w rarely 0.8 w. i am unable to upload the power txt file here . https://pastebin.com/p0F35ySB power.txt
image after clean boot.

 

@Docktor - Interesting that Package C8 stops working after resuming from Sleep. I had a hunch that it might be related to this. I will let you know if I see anything unusual in your powercfg report.

You could try confirming this problem by using HWiNFO. You could also ask your laptop manufacturer about this problem but usually the person you contact will not have a clue as to what you are talking about. They will probably brush your complaint off and tell you that they do not support 3rd party monitoring tools. Trying to get to the bottom of a problem like this could be very difficult. Sad that Intel creates power saving technologies that are not being used properly by manufacturers.

 

@unclewebb yes this is a weird interesting problem. HWiNFO tells the same story so its confirmed resuming from sleep c8 , c7 stop working.
i dont think the laptop manufacturer has anything to with this issue because c8 was working before , i noticed this issue after i used the unlocking method for advanced power options posted by vasudev . So this problem might be because of that, is there anyway to undo that without reset or clean install of windows?
i checked the c states because there were options for core parking in the unlocked options to see how they affect them. i did reset the options to default when i noticed this problem.
hopefully you can find a solution to this .

 

they are on default settings i was asking about undoing the unlock of advanced power options.

 

I have the following issue, that TS is force closed (by Windows mostly) with a high chance if I remove AC and go on battery. I have set up to start TS via the Windows task planer. I cant see anything wrong there, everything which could indicate to force close it if not on AC is not selected.

The trigger is on user login, high privileges, conditions everything is deactivated (no just run on AC or anything).

Anyone else have this issue or knows how to solve it?

 

Have the same issue couple of times. I tried disabling Power throttling on TS under battery saver and allowed TS to run in background. That decreased the occurrence of force closing TS. But it happens intermittently whilst switching from battery to plugged in state or vice-versa.

 

@maffle - ThrottleStop shutting down when switching to battery power is a known issue. It is caused by your Task Scheduler settings. Follow the guide in the second post of this thread. The Task Scheduler has some options that are grayed out. Users think these settings are not active when in fact they are active and need to be disabled. The guide tells you how to do this.

http://forum.notebookreview.com/threads/the-throttlestop-guide.531329/#post-6865107

I do not own a laptop with C8 capabilities for testing purposes so there is no way for me to find out what setting screwed up your package C8. Are you sure this was not a problem before you applied those power plan changes? Perhaps this problem was already there but you did not notice it before.

 

I have the same problem. I disabled power throttling for TS and to a certain extent it fixed the issue.
Scenario: Constant switching to AC/Battery say twice or thrice consecutively could force close TS abruptly.

 

@unclewebb I am positive this issue was not present before . If you can't find anything maybe windows reset can fix this because afaik unlocking the advanced power options is the suspect.
so should i wait or go ahead with the reset ? this is not a system breaking bug/issue so i can wait.

Also again is there a way to undo the unlocked advanced power options

 

@unclewebb I stated above, that all scheduler settings are correct. It still happens. There is nothing selected like "just run on AC" or anything. All settings look 100% identical as in the guide you posted and yet, it is forced closed with high chance if I go battery. If I then look into the scheduler, it says something as status "task has unexpected ended". You should really write a proper Windows service. But this issue is not related to a wrong setting it seems. Wonder if this is some typical Windows 10 bug... or could it be, that TS just crashes for some reasons the moment I unplug AC?

 

Have a look at the Task Scheduler Conditions tab. Make sure all options are clear. Even the grayed out options have to be clear.

When properly setup, the Task Scheduler works 100% perfectly for me. I run the same instance of ThrottleStop for months at a time without it ever crashing or being forced to exit by Windows.

I was recently given some new Windows startup code from the W1zzard who runs TechPowerUp. My long term plan is to add this new code to ThrottleStop later this year so users will not have to mess around with the Task Scheduler.

 

@unclewebb @Mr. Fox @Khenglish


Am I misunderstanding something or did someone manage to undervolt an Ivy CPU?

Because this tool claims to use similar methods to throttlestop but only works for 3rd gen (Ivy) CPUs

https://github.com/tiziw/iuvolt



And this tool claims it works on 2nd gen or older: https://wiki.archlinux.org/index.php/PHC

 

oops

 

Yes PHC with patched kernel for UV'ing works using TLP.
You can ping the author @tiziw

 

Then can throttlestop be made to work with older gen cpus?

 

MSR 0x150 did not exist in the 3rd Gen Core i or earlier CPUs. That MSR is only available on 4th Gen Core i. His documentation needs to be updated.

If you have a 3rd Gen Core i handy, run RWEverything and try to read that register. RWEverything is not very forgiving so be prepared for a BSOD if it cannot find that register.

Edit - I went to the PHC website and read the Success Stories section. Lots of people under volting Core 2 Duo based CPUs but I could not see anything posted for Core i 1st Gen or Core i 2nd Gen.

 

The register don't exist for Ivy and below right?

Seems like author for the tools I linked are mistaken.

 

There's supposed to be an option to flip the "additional turbo voltage" to be negative, but that's absent on Clevo systems. Setting an absolute fixed voltage should also work but is also missing.

What I would really like to see is being able to set a fixed voltage. Right now I can only do that by editing my bios to have all C-states higher than C1 disabled. When I do this I can clock my CPU over 4.4 GHz without extra voltage, but with C3 and higher enabled the CPU will crash on the upward clock transition forcing me to run at 4.2 GHz for stability. Throttlestop's C-state options don't have any effect. Any modified power plan I've tried in Windows also disables C1, which means the CPU burns 70W+ at all times, which is not really acceptable.

 

Does some laptop have the ability to use a negative "additional turbo voltage"?

 

I think it is only functional on desktop systems.

 

aww

 

Well I found Unclewebb answered my question over here on techpowerup:

https://www.techpowerup.com/forums/threads/cpu-throttling-at-high-performance-settings.221607/page-2

Lack of to multiplier due to no C3+ is no problem on an unlocked CPU. 200+ MHz for free.

 

I undervolted my i7 8750h (-155) and temps improved drastically but my cpu seems to throttle when it reaches 80c.

I thoight they are supposed to go up to mid/high 90s before they throttle.

Even weirder i was playing cs go before i undervolted and my temps would go as high as 98c before it throttled. So i don't know what's causing this throttle to 80c?

 

hahaha.

ThrottleStop works great on my older laptops, including my Core 2 Duo MacBook Pro. Gives these laptops yet another life. . .

 

The throttle point is stored in the Bios under Thermal configuration. Usually the setting "TCC offset" what changes this from the default point of 100C.
100C is the intel designed point. Cancer Bioses often set an offset of 20, dropping this to 80C.

You would have to find out if you can view the bios menus in something like AMIBCP 5.02 or newer, and if you can, if you cant manually unlock menus with FPTW64 (to dump the bios first-WARNING: NEVER EVER attempt to run AMIBCP to do anything more than View the bios options on the bios downloaded from the OEM's website; if you're trying to unlock menus, you have to create your own dump of your system bios manually with FPTW64), run AMIBCP 5.02 to set all of the advanced menus to "Supervisor" and then reflash the APTIO capsule wtih FPTW64, if you cannot do this, if the option of "TCC offset/index" even exists in your bios and if you can view it, you may be able to create a dump of the bios setup capsule with UEFITool, then open that dump you made with Universal IFR extractor. Then you can save that as a text file, then look for the hex offset string next to the "TCC" option and find it (example: 0xA78). Then you can try to disable secure boot and boot to "RU" EFI prompt, and manually change the TCC value to 0 that way.

If you can't do this you're screwed. .

 

Did you under volt the CPU Core and CPU Cache? Your CPU might be throttling at 80°C now but is that the real reason it is throttling? Most throttling is power limit based. Perhaps the reason your CPU is throttling is because you have reached one of the built in power limits. It also might have been forced into TDP Level 1 so now your CPU is being limited to 35 Watts instead of its rated 45 Watts. Manufacturers have lots of tricks up their sleeves to make their laptops run miserably.

Open up the Limit Reasons box in ThrottleStop, click on the headings to clear the previous info and then go play a game and see if those boxes light up while playing. Boxes in red show throttling in progress and boxes in yellow indicate that throttling was previously taking place. Here is an example of my 4700MQ being throttling to its 47 Watt TDP limit (PL1).



I would run the Log File option in ThrottleStop but before doing this, go into the Options window and enable "Add Limit Reasons to Log File" and also enable GPU temperature monitoring if your GPU is supported. This extra information might explain better why your CPU is throttling. Posting lots of screenshots and log file data makes me happy. www.pastebin.com works well for log files.