The ThrottleStop Guide

Try removing the keyboard+bottom cover and run it. With any luck you can pinpoint the exact culprit and either replace yourself or have a shop do it for you. Easiest is to 'home in' and touch the various ICs until you've found the one that's silenced by this. Did the same for an adapter that was noisy under load; looks slightly used now, but at least it's quiet.

 

When I choose multiplier this obviously boosts all cores... My laptop doesn't have turbo-boost option enabled and can't via bios.. (limited bios)
Now I wish to go further then x36 multiplier (which is the normal turbo boost for cores for my i5 6600)

Which means x39 for 1 core and 38 for 2 cores.
Can I do this with this program? and if not any other way.?

 

http://ark.intel.com/products/88188/Intel-Core-i5-6600-Processor-6M-Cache-up-to-3_90-GHz

The base frequency for a Core i5-6600 is 3.3 GHz. If your CPU is running faster than this then Turbo Boost is already enabled. After you have booted up and Windows has had a chance to settle down, open up TS Bench and run a 1 Thread test. Post a screenshot of the main ThrottleStop window while the 32M test is in the middle of running. What multiplier and MHz does it show?

In the FIVR window, ThrottleStop should show 39, 38, 37, 36 for the 1 Core Active, 2 Cores Active, 3 Cores Active, 4 Cores Active. Is that OK?

Edit - Also make sure that the Windows Power Options - Maximum processor state is set to 100%. If this is set to 99% or less, Turbo Boost will be disabled.

http://i.imgur.com/PzUuxhn.png

 

http://imgur.com/VUKbEO8

It shows me 37.5 on all cores lol 3800 MHz on all 4 cores.. this is higher then it should be I think?? 36 on all 4 cores should be normal?
Yes and set power options like you said.

 

Update on the Skylake mobile support. My SP4 can undervolt as low as -137mV before crazy things start happening -- but that's only "stable" plugged in because the device goes into idle afterward and the voltage at the idle speed (0.46GHz according to task manager -- TS's Skylake clock report is still wrong and always shows the max with very little fluctuation) is apparently insufficient after the heat of a benchmark.

I can set it manually (static) to 0.8975 and it does okay. This is some crazy undervolting -- I've gotten it almost all the way through the 1000M benchmark without the fan turning on, but near the end of the TS benchmark (around 80%), the temps breach 60C and the fan turns on. I'm frustratingly close to being rid of the fan. Contrast this to the fan turning on before 7% (as is the case without any undervolting).

Anyway, a few more reports concerning TS's functionality with the Skylake U processor:
- the new VR settings (checkboxes) seem to have no effect on temperature, performance, or stability of the 6650U in my tests. I guess this isn't unexpected.
- The cache ratio can go as low as 4, but entering anything into the box causes it to limit the minimum cache ratio to 8.
- "Edit - Also make sure that the Windows Power Options - Maximum processor state is set to 100%. If this is set to 99% or less, Turbo Boost will be disabled." -- Connected Standy devices have no such option. There is no way for the user to control the maximum or minimum processor state in Windows Power Options. If there were, a lot of my voltage problems would be fixed because I could define a minimum greater than the multiplier at which my CPU's voltage destabilizes at idle.
- The skylake is 2-3x faster than the atom x5-8500 in my other tablet. With the 32M benchmark, the tablet gets 127s, 70s, 48s for 1,2, and 4 threads. The SP4 gets 57, 29, 16.5s respectively. Interesting how the two architectures (cherry trail and skylake) differ so dramatically in their performance.

 

If the LoadTester does not easily elimnate the high pitch noise, are there any options left?

 

I fear it has come down to somehow disassembling my laptop (not exactly one of the most easy ones to disassemble) and putting glue over the capasitors making the noise. I can't use the god damn thing with this whining noise

 

Hello everybody.
I have got a problem.
My specs are:Laptop Dell inspiron 5737 (i7 4500U). The MAX frequency is 3.0 (1 core enabled) 2.7 (2 core enabled). Max non turbo boost speed is 1.8.
After clean reinstal Windows 10 x64 my CPU frequency drops in "High perfomance" Windows mode from 2.7 to 2.0 Ghz (without TDP Throttle). The min and max CPU State in (High perfomance mod) are 100%.
I don`t have this problem when i use "Balanced" Windows mode (Frequecy normaly jumps from 800 mhz to 2.7 Ghz). And i didn`t have such problen in past.

All drivers are already installed correctly.
I think that the EDP is the problem.
I attached some screenshots below.
Could you help me to find solution?
Thanks a lot.

Spoiler: Clean start

http://imgur.com/a/3LL4F

Spoiler: High Perfomance

http://imgur.com/Mio8ZmT

Spoiler: Balanced

http://imgur.com/uZkbU4w

Sorry for my English)

 

Welcome to NBR. It looks like EDP throttling is the problem.

http://forum.notebookreview.com/threads/tdp-and-power-limiting-haswell.766743/

Dufus did some thorough testing and found out that EDP can be triggered when the CPU reaches the current limit but it can also be triggered when the CPU reaches the PL3 power limit. I started adding control of PL3 to ThrottleStop but I didn't know of any hardware that is using PL3 for throttling purposes so I put this back on the things to do list. Maybe it is time to resume work on adding PL3 control to ThrottleStop.

In the Balanced profile, can you access the Control Panel - Power Options - Minimum processor state? This is usually set to 5%. Try setting this to 100% and see if EDP lights up red in the Limit Reasons tool that Dufus wrote. This might be a Windows bug.

 

Hello there.

First of all, I would like to thank you for such a great tool, which is truly one-of-a-kind. I am using it all the time and it works great.

I've seen that you have had some Bay Trail atom specimens information, but just in case, I am including the MSR Finder info of my Z3735F, as I'm not sure whether you got the info from a newer Bay Trail processor (Q1'14 vs Q3'13).

http://pastebin.com/hz1dDEwQ

Anyways, there is just one thing, which I would like to ask: is it possible to change the VID of different P steps? As currently, it seems that I can change only the turbo VID, which is set very generously - going from 0.560 down to 0.500 is fully stable. I was just thinking that I could probably save some more power this way.

Anyways, thank you once again.

 

I am sorry for spamming this thread with my issue but it seems like the only place where there's knowledgeable people regarding the subject.

I found out why my CPU does not whine on battery: the c-states are not being used at all! Atleast according to Throttlestop, under "package c-states", they're all on 0. This would also explain the subpar life that my battery were performing, I suppose.

I decided to give Windows 10 another go in a dual boot, and despite C6 state being fully used, both on battery and AC, there does not seem to be a high pitch audible? Not yet atleast, but I don't see why it would change. I monitored the clock and c-states to make sure it was not because of Windows doing something in the backgrund, and seemingly it is not. I do not understand how diference in an operating system can do this, I thought it was hardware related at most?

 

Unfortunately, no. The original goal of ThrottleStop was maximum performance while plugged in. Saving power while unplugged was never a major consideration so it does not have the intermediate P State feature that RM Clock used to have. After reading Power Optimization - A Reality Check, I decided to ignore the intermediate P States.

http://www.cs.berkeley.edu/~krioukov/realityCheck.pdf

Modern CPUs should be spending only a very tiny percentage of time at the intermediate P states. If a CPU has something to do, it should do it as fast as possible in the highest P State where it is most efficient. As soon as it is done, it should return to the lowest P State, enter one of the low power C States and save power that way. I might be wrong but I just don't see any opportunity for significant power savings based on how modern CPUs and modern operating systems are designed to work.

Have you tried using the ThrottleStop - Power Saver feature? I am not sure what the savings will be but this should let your CPU switch to the minimum FID and VID whenever the CPU is idle.

I assume you mean the Package C States are not being used. Are the individual Core C States being used on battery power? Post some screenshots so I can have a look. Originally when I switched from Windows 8.1 to Windows 10, I noticed a lot of background activity going on which was interfering with the CPU using the lower C States. After a couple of days I finally got fed up so I neutered W10 and forced it to settle down until C State residency time was about the same as previous OS versions.

The C State data that ThrottleStop reports is very accurate but it is hardly talked about in the forums so spam away. I think TS is a great tool for hunting down poorly programmed drivers and other crapware running on your system. Different operating systems use different drivers and one bad driver can completely change C State residency time. The Elan track pad driver that I was using while on Windows 8 was a disaster of poor programming.

 

I should've taken a screenshot. Seems liek Wndows 10 wont let me boot back to 7.

Package c-states seemed all to be on "0.0" when on battery, afaik. No whining noise.. Plugged in AC and C6 especially went up in use, and the whining could be heard.

Battery life improved immensely on Windows 10. I did a test on Windows 7 the other day (because the bad battery life bothered me) and the result with min brightness, no wlan and idle was around 3,40 hours at best. Now on Win10 its around doubled, maybe a bit under - 6½ hours or so. I dont know what I did on Windows 7 to kill the power saving states when on battery. But like I said, here on Windows 10 the hissing is close to inaudible. So, do I wanna change that? If I "neuter" it the same way you did, would I be back to the whine sound, likely? According to fan noise, cpu percent usage and so on, there doesn't seem to be anything going on?

 

Thank You very much for your answer.
I have tested the "Balanced" mode with minimum CPU State 100%. The CPU frequency jumps from 2.0 GHz to 2.7 GHz. EDP lights are yelow . Why i didn`t have this problem on Windows 8.1 and Windows 10 in past? Could this be due to new Windows 10 updates?

 

@Kocane - Sounds like you have a couple of good reasons to stick with Windows 10. Better battery run time and less whine. If everything is OK now, no need to keep fussing with it.

@LuckyDi - When Limit Reasons shows a box in Yellow, that just shows that some EDP throttling did occur since you booted up or since this was last reset. There might have only been a few milliseconds of throttling while booting up. You can click on the headings in Limit Reasons and it will reset the CPU so it should show all black. After that, go play a game and then you will have a record if Limit Reasons lights up again.

It is impossible to say if a Windows update screwed up your system. The purpose of some updates are not well documented. Microsoft seems to be sliding in more questionable updates under the category of Critical. If I choose to continue using Windows 7, stop hitting me over the head with W10 messages that are as annoying as some viruses. I have Windows 10 on my laptop. I turned off all updates so I don't have to waste endless time trying to figure out what just went wrong with my computer.

 

I feel for you, Im assuming you've played with voltages, intel IXU already and loads.
I'd honestly suggest a possible bios update, or at the least a chipset driver update.

All I can think of is to disassemble, pinpoint the noise, and muffle it with a nice piece of foam, it sounds like it's crystal or voltage regulator module related. If you don't have foam around I could always mail you some in a letter. Worst case scenario treat yourself to nice earbuds to use while laptoping.

*I didn't read everything sorry, sounds good that the new windows seems to handle c states better.....makes me feel like the chipset drivers were still the original culprit.

 

finally a program running on my new notebook with skylake 6700hq Processor!! Much much appreciated work from you dude! I almost gave it up searching for a program undervolting my notebook for better batterylife.
Now digging into this thread, i found out that skylake CPUs are not yet supportet fully, but some thinks my function yet as i found out trying your program. First setting a static Voltage for the cpu does work, but only to current minimum VID that the CPU programmed to run with. That means the idle VID of my CPU is about 0.85v and setting anything lower then that will not be applied. Setting anything higher then that works easily. Same under load conditions: When fireing Prime95 up, multiplier goes to the disired Level bringing up a VID of 0.97 Volt at 2.6ghz. (i disabled TurboBoost to save some Battery). Setting anything lower then 0.97v doesnt work, setting anything higher then 0.97 volt works perfectly.
Now sadly i just wanted the opposite to squeeze some more Battery out of my Dell XPS 15 with the 6700HQ CPU.
Wanted to ask u if theres still some trick to make UV work or is this not working yet to the current state of throttlestop?
If i can provide u with any infos, pls tell me i will do it as fast as possible. Happy about your work very much.

Thx again and cheers to all

 

dudes i just realized that after a second installation of Intel XTU, undervolt in adaptive mode completly works perfect! I am able to undervolt my 6700HQ to a massive -0.1875v idling at 0.685v and running at a stuning 0.80v at full load @ 2.6ghz all cores HT included. Power drops from 34watts to a nicely 22Watts and Temperature drops from 70° to 52 degrees. This is massive So my question is solved at that point

 

I'm a big fan of Windows 7, not so much 10, and despite 10 being more kind on my laptop in terms of someone managing my CPU power saving states (less highpitch noise) I decided I wanted to try something in Windows 7. I had previously accidentally done so c-states in Windows 7 were not being used while on battery. This was due to the settings I had made in my power settings plan, under processor. With a certain registry edit you can enable some hidden functions, by running these line by line in CMD or just as one .bat file:

• powercfg -attributes SUB_PROCESSOR 06cadf0e-64ed-448a-8927-ce7bf90eb35d -ATTRIB_HIDE
• powercfg -attributes SUB_PROCESSOR 12a0ab44-fe28-4fa9-b3bd-4b64f44960a6 -ATTRIB_HIDE
• powercfg -attributes SUB_PROCESSOR 40fbefc7-2e9d-4d25-a185-0cfd8574bac6 -ATTRIB_HIDE
• powercfg -attributes SUB_PROCESSOR 4b92d758-5a24-4851-a470-815d78aee119 -ATTRIB_HIDE
• powercfg -attributes SUB_PROCESSOR 7b224883-b3cc-4d79-819f-8374152cbe7c -ATTRIB_HIDE
• powercfg -attributes SUB_PROCESSOR 943c8cb6-6f93-4227-ad87-e9a3feec08d1 -ATTRIB_HIDE

I'm actually just using one of them, but I don't know which of the 6 tweaks is the one and I don't mind having more options

What's interesting is that when on battery, by default, the powersaving state "C7" is mainly, if not exclusively, being used... and this c-state does not (atleast with my pc) omit any noticeable whine. So I thought, if I just did like I had accidentally done previously, disabled these c-states completely (or the packaged c-states anyway?) but this time on while plugged in, I could avoid C6.

This is "before" while plugged in:


While on battery:


By using the new power setting tweaks, I went into my power plan and set "processor idle promote threshold" to 100% on plugged in and set the maximum processor state, while plugged in, to 100% as well. Doing this elimanted the whine and set all c-states to 0.0, according to throttlestop:



In contrast to the tweak (also including in the 6 registry tweaks above) that allows to disable idle completely, this does not really seem to affect temperatures all that much. It makes the PC be less idle on lower clocks but while plugged in - who cares? Atleast it's better than the whine.

and ofcourse, one can use dear throttlestop to create some profiles, like an internet profile that uses the "power saver" option which definitely pushes clocks down. How much an affect it has on power usage or heat, I'm not really sure

 

@schidddy - Nice voltage numbers. I have heard that when using ThrottleStop on Skylake, you need to adjust both the core voltage and cache voltage equally or it will not work. No hardware yet to play with this.

@intruder16 - I am still running Windows 10 10240 on my Y510P. No issues.

I originally planned to add a variety of options to control PL3 but then I thought about it and asked myself, "Who in their right mind wants to throttle their computer using PL3?" Answer: Nobody.

Based on this I decided to add a simple feature. The Disable and Lock PL3 option. If you have a device that uses this throttling method, checking this box will set this register to zero so it is not used and then it will lock it at zero so you will not have to worry about your laptop or tablet setting it. Final testing in progress. Hopefully it will be ready in a few hours.

 

ThrottleStop 8.00 beta 7
https://www.sendspace.com/file/i84h7g

New Features:
- added Power Limit 3 (PL3) Lock and Disable feature.
- removed Limit Reasons in the log file for Skylake CPUs.

Here is an example of how the new Power Limit 3 feature works. On my Y510P, when plugged in, the PL3 register is set to zero which means that Lenovo is not using PL3 to throttle the CPU.



If you open up the TPL window and you see some odd ball looking number in that box, that means your device is using the PL3 register for throttling purposes.



If you see EDP lighting up in the Limit Reasons program, that is another sign that your computer might be using PL3.

And now for the easy fix. If this register is not locked, click on the Disable and Lock PL3 box and then push the Apply button. The number in the box should change to 80000000. The 8 signifies the lock bit and the rest of the PL3 register is set to zero so this register should no longer be a thorn in your backside.



I am not sure what devices are using PL3. I am thinking that this might be a more popular throttling method in low power laptops and tablets but who knows what manufacturers are up to. Post a pic if you see something hiding in the PL3 register.

 

Processor Performance Increase Threshold
powercfg -attributes SUB_PROCESSOR 06cadf0e-64ed-448a-8927-ce7bf90eb35d -ATTRIB_HIDE

Processor Performance Decrease Threshold
powercfg -attributes SUB_PROCESSOR 12a0ab44-fe28-4fa9-b3bd-4b64f44960a6 -ATTRIB_HIDE

Processor Performance Decrease Policy
powercfg -attributes SUB_PROCESSOR 40fbefc7-2e9d-4d25-a185-0cfd8574bac6 -ATTRIB_HIDE

Processor Idle Demote Threshold
powercfg -attributes SUB_PROCESSOR 4b92d758-5a24-4851-a470-815d78aee119 -ATTRIB_HIDE

Idle Promote
powercfg -attributes SUB_PROCESSOR 7b224883-b3cc-4d79-819f-8374152cbe7c -ATTRIB_HIDE
https://msdn.microsoft.com/en-us/library/windows/hardware/mt422914(v=vs.85).aspx

CPOVERUTIL
powercfg -attributes SUB_PROCESSOR 943c8cb6-6f93-4227-ad87-e9a3feec08d1 -ATTRIB_HIDE
https://msdn.microsoft.com/en-us/library/windows/hardware/mt422922(v=vs.85).aspx


@Kocane - I just updated your list with the proper name for each one. I think Idle Promote should give you some control over the C States in Windows 10. Thanks for the info.

 

Hey Unclewebb, I was recently helping a friend use throttlestop to try and overclock his 5820K, and I discovered some interesting things.

Apparently, if the BIOS on an ASUS X99 board is limiting the CPU speed (in this case, he was stuck at 2.9GHz for some odd reason) then throttlestop's set multiplier and/or overclock (in the FIVR menu) features don't work. But the voltage control worked, which meant TS was working to some degree.

Also, it seemed to read the power usage incorrectly. Running TSBench it claimed he was using exactly 4W and when idle he was supposedly using under 1W. I saw it via skype video, so I can confirm actually seeing it myself (I was telling him what to do). I don't know if I could get him to be bothered to submit a log file or anything as he's usually a bit hassled by stuff like that, but I figured I would mention it. If any other ASUS ROG Haswell-E users on the forum want to help out assuming they have the same issue, that might be okay. I believe @johnksss has one such system, and I'm not sure but @TBoneSan might too. As for if they're willing to help, that's up to them.

 

Thanks for clearing up what each registry tweak does. The idle promote also gives control in Windows 7 but it seems like I have to set the minimum proc state to 100% otherwise there's no difference.

 

I observed some odd behavior too. I've reluctantly not been using it. Voltage control didn't seem to work on profiles and clocks wouldn't always stick too. I'll have another look when I get home tonight and link some logs or whatever's necessary if things act screwie.

 

Yeah. Also, something I forgot to mention too; if the PC is rebooted while Throttlestop is running and it has adjusted voltage, that voltage sticks in the sBIOS readouts. Only observed on X99 so far.

 

Interesting - I'll get back to you on that.

UPDATE: Just out of interest. What was selected here?



I've installed TS8 b7 - I'll monitor it more over the weekend.

I've noticed so far that clocks can't be increased beyond what's been set in the BIOS.
Ie. if my BIOS is set to 4.4 but want TS to take it beyond that to 4.5 + on a performance profile.
- this may be by design, I don't remember TS doing this with my 3920xm.
I guess I can easily enough do things the other way round if need be.
Ie. 4.5 in the BIOS and have the TS default profile at 4.4 upon startup .

I'm happy to report that voltage controls are working a treat, and stick to each of the profiles. I'm stoked to have that control again.

 

The same option you have @TBoneSan. Not saving changes to the .ini file.

My BIOS has my CPU multipliers at 3.5GHz though, but if I set them to 3.8GHz in throttlestop they'll increase. So it has to be a X99 deal. I'm not sure about the 39x0XM chips however. I have however tried TS8 beta 2 and beta 5 with a desktop 4790K and it has been capable of overclocking properly.

And yes, voltage controls were working in my test, though the rebooting and having voltage stick was a weird (but once-off; we didn't try it again) one.

 

These Skylake CPUs undervolt crazy amounts. I am able to run full load while gaming (haven't tested anything lower yet) @ -160mV. I've been gaming and using the laptop for weeks with this undervolt and not a single hiccup or crash and it holds the maximum clock speeds. I just set -165mV and will probably continue to go further until I get a crash. That undervolt shaves off some serious heat and comes at zero performance penalty.

 

Tried undervolting a desktop Haswell once, a surprising -220mV which meant it operated at lower voltage and power than comparable dual core mobile ULV's. :/

Don't forget PL4 and RATL in the ever growing ways to throttle


FYI with powercfg you can use the aliases where provided (powercfg -aliases), unless one is good at remembering GUIDs such as

"powercfg -attributes SUB_PROCESSOR CPOVERUTIL -ATTRIB_HIDE"

 

I'm at - 0.1925v now meaning a full load voltage at 0.775v now haha, thats some serious ****, working with the laptop now lasts much longer Lesser voltage will make cores stop working on Prime on longer runs. So enough OT now thx for this thread and keep up the good work guys!

 

Hello @unclewebb

Is there a way, with any TS vesion to stock my i5 520m to it max multiplier and undervolt it?

Thanks and greetings from Peru

 

The Save Voltages section is a safety feature that lets you decide when your settings should be saved for future use. XTU used to irritate me because it had a bad habit of dumping all of my voltage settings whenever I tried to adjust the voltage a little too much. For ThrottleStop, the goal was to give users more control when experimenting with voltages. When the first option is enabled, you can play with the voltages and no matter what happens, those voltages will not be saved to the ThrottleStop configuration (INI) file. After a crash, when you start ThrottleStop, it will use the last safe voltages that you saved so everything should be OK. You won't have to worry about getting stuck in an infinite boot loop because ThrottleStop saved some unsafe voltages and it keeps trying to use those over and over again.

The second option will only save the new voltages after ThrottleStop safely exits. That way you can change voltages, go do some bench testing and if your computer doesn't crap out, exit ThrottleStop and your new voltages will be saved to the INI file for future use. Sometimes after adjusting voltages, you click on OK and your computer freezes up solid. With this option, no worries. ThrottleStop did not safely exit so it assumes that something went wrong and those voltages that you were checking out must have been bad so they were not saved.

Final option is for the experts. The people that know what they are doing and don't want any damn training wheels. ThrottleStop will immediately save whatever voltage changes you make to the configuration file for future use. If your computer locks up, it will probably crash if you try to run ThrottleStop again because it will try to run the stupid voltages you previously entered. Nothing but grief when things go wrong. You would need to delete the INI config file and start from scratch if you use this option and things go bad. Not sure why the experts use this setting. I hope that was the question you were asking.

If you have any problems, post some more screenshots. ThrottleStop does not have access to some registers that XTU has access to. There are various ways to work around some of these issues. Sounds like you found a way that works.

@clasho - It is not possible to under volt a first generation mobile Core i CPU. You can adjust the turbo power limits on an i5-520M but you probably will not need to do that. Run the TS Bench test or Prime95 and if your computer does not run at its rated speed, you can try increasing the power limit in the TPL window. When both cores are active during a stress test, the maximum multiplier should be showing 20.00.

PL4 sounds familiar but RATL is new to me. I don't think Lisa is going to be very happy with you or me if I go much further with TS. Of course, there are still one or two more things I have been thinking about.

Here is what my W10 laptop shows for powercfg -aliases in the SUB_PROCESSOR category.

Code:

54533251-82be-4824-96c1-47b60b740d00  SUB_PROCESSOR
0cc5b647-c1df-4637-891a-dec35c318583    CPMINCORES
4b92d758-5a24-4851-a470-815d78aee119    IDLEDEMOTE
5d76a2ca-e8c0-402f-a133-2158492d58ad    IDLEDISABLE
7b224883-b3cc-4d79-819f-8374152cbe7c    IDLEPROMOTE
bc5038f7-23e0-4960-96da-33abaf5935ec    PROCTHROTTLEMAX
893dee8e-2bef-41e0-89c6-b55d0929964c    PROCTHROTTLEMIN
94d3a615-a899-4ac5-ae2b-e4d8f634367f    SYSCOOLPOL

 

Ah, maybe that explains why I have trouble pm'ing info.

I'm sure you can guess RATL stands for Rolling Average Temperature Limit and similar to RAPL for power limiting.

W10 hiding stuff? :/ You could try adding the hidden option "powercfg -aliasesh"

 

My PM settings have been adjusted. Adding some of those hidden powercfg options directly to ThrottleStop might be interesting. There goes 2016.

 

You say "there goes 2016" but let's be honest, you're cackling like a mad scientist inside right?

 

When adjusting the C7 interrupt response to something like 1, it seems like some cpu whine can also be stopped but what exactly does it do? If we look at package c-state percent, they're all except C2 at 0.0 but in core c-state percent, they're all using C7 very much. What does this mean to power saving?

 

Maybe.

@Kocane - Your individual cores are still using the C7 low power C State but the entire CPU package is being blocked from going into any of the C States other than C2. I think doing this has been known to reduce whine while maintaining decent power consumption. In theory, you should get slightly less power consumption if you allow the CPU to use the C6 Package C State compared to the C2 Package C State but the difference is not huge as far as I know.

 

Definitely did seem to kill off the worst whine. I'm wondering though; when exactly does a state like C6 come in handy? I mean, from what I can read it is quite deep sleep which should reduce power consumption to almost non. But wouldn't the CPU jump out of this high power saving the moment it's being used?

I'm trying to do some battery testing but it's mainly just looking at Windows and BatteryBar estimates. Windows 10 did say 5-6 hours from the start and Windows 7 around 4-5 but I don't know how much to make of that. I'll have to do a full test of 100% -> 0% with lowest brightness and wifi off. It'll be interesting to know

Also, it'll definitely be cool if you make it so Throttlestop can manage the hidden Windows proc power settings, looking forward to that!

EDIT: After some quick testing on lowest brightness, wifi off and idling I must conclude that the difference between C7 interrupt response set to 1 and 10 is that with 10 (package c-state being used) is like 25% percent. The discharge rate with 1 is 8300-8500 mW where-as with 1 it's like 6000-6300 mW or something like that... but yeah, that's really completely idle so I don't know how much of a "real world" battery difference it has. This test is basically just using C7 and not using C7?

It could be cool if it could be done so that when the PC is actually idle it automatically switches to a throttlestop profile with a different C7 interrupt response setting and switches back when the PC is being used.

 

At the moment, you can only do profile switching based on CPU temperature. You should be able to set up 2 different profiles and then switch based on temperature. When your CPU is mostly idle, it is going to be running at a lower temperature so this might more or less work. Perhaps someday I will add profile switching based on load.

Edit - Just did some quick testing with Battery Bar. When I adjust C7 IRL from 12 to 5, my 4700MQ goes from spending 84% in the Package C6 state to about 95% in the Package C2 state. Power consumption goes from ~11 Watts to ~16 Watts. When running on battery power, using the package C States looks like a good idea. Would need to run a movie or something like that for an hour to see how meaningful the Battery Bar data is. Percentage wise, that is a big difference and should make a significant difference in run time.

 

Hi, your tool is amazing. Now I can finally keep my Core-i5 4200U from getting too hot and loud.

Just one question: How can I set profile switching based on CPU Temperature? I want a profile for low temperatures and a profile for high temperatures which disables the turbo boost to prevent CPU from getting too hot when running on 100% load for a longer period of time.

And for me the program adjusts CPU settings without me clicking on "Turn On".

 

I always like hearing stuff like that. ThrottleStop is a little complicated at first but once you understand the program, a few simple mouse clicks can make a big difference to how a laptop runs.

The Turn On / Turn Off feature only applies to the Clock Modulation, Chipset Clock Modulation and the Set Multiplier values. Many newer laptops do not need to use any of these features. They can be fully controlled without having to click on this. My suggestion has always been to only use the ThrottleStop features that you need to use. There is no use clicking on some of these items if your laptop does not use these throttling methods.

Intel CPUs are rated to run reliably right up until they reach the thermal throttling temperature which is 100°C for your 4200U. You do not need to protect it because they are well designed to protect themselves from any damage due to heat.

If you still really want to set up 2 profiles then use the Alarm feature.



A DTS setting of 30 means your CPU will switch profiles when it is 30 degrees away from the thermal throttling temperature. In the above example, when your CPU reaches 70°C (100-30), it will switch from whatever profile it is on to profile 2. I typically use profile 1 as my main profile so it would switch from 1 to 2. Once the temperature got back under 70°C, it would switch back from profile 2 to profile 1. This might be a useful feature for some users but for me, I just let my Intel CPU take care of itself. Properly designed thermal throttling at the hardware level works a lot better and faster than doing this in software.

 

Thx for your answer!

The original issue for me was this: The 4200U has a strange thermal control build in. It runs on turbo (multiplier of 23 to 26) nearly all the time. The temps are usually between 60 and 70 degrees celcius. But when under 100% load for some minutes, it starts to throttle to 800 MHz for 5 min until it cools down, which is not necessary. I can see it on the temperatures. I don't know how to deactivate this throttle. May be it is set by the manufacturer (notebook is a VAIO Flip 14 refresh).

So my attempt was to reduce the core voltage by 60 mV which worked quite well. After some tests I discovered that with current settings I can only run on 100% continuous load without throttle when I disable the turbo. CPU runs then with a multiplier of 16 and temps are little below 70 °C (with constant 100% load). "BD PROCHOT" is already disabled.

I know you have a widget to show throttle reasons. I haven't used it yet.

Anyways my next upgrade for my stationary computer will be intel as well and I can't wait to optimize it with ThrottleStop.

 

Did you ever program a way to change profile via commandline/shortcut?

 

How to know the default VCCIN of my CPU?

Sent from my D6503 using Tapatalk

 

@Cursy - The 4200U is a 15 Watt CPU. Short term they can run a little beyond 15 Watts but long term, no one has found a way to have them run beyond 15 Watts. The throttling you are seeing is usually not temperature related; it is power related. The U CPUs also have another feature called TDP Level. The ThrottleStop TPL button will open up a window and show you what TDP Levels your CPU supports. When one of these CPUs drops into a low TDP Level, it will be limited to far less than 15 Watts. Usually somewhere around half that amount. When it goes into this low power level, the multiplier is also limited to 8 so your CPU is left limping along at a hair under 800 MHz.

Sadly, there is nothing ThrottleStop can do about this. You can use ThrottleStop to try and lock a CPU register into high TDP mode but there is also a secondary register in the chipset that ThrottleStop does not have access to. This secondary register might also be locked. All you can do is reduce voltage and CPU speed so your CPU does not go over the cliff and end up in low power TDP mode. Each individual manufacturer gets to choose when to trigger low TDP mode. Too much control is with the manufacturer so a 4200U in one device can run completely different than the same 4200U in a different device. No way for a consumer to know what they are getting until after they buy a device, take it home and do some thorough testing and usually by then it is too late to take it back. That's why I do not recommend these things. Post a TPL screenshot so I can have a look.

I have not done any command line type programming so I have not looked into this. Maybe someday.

I do not know of any way to read the default VCCIN from a 4th Gen Core i processor. For my 4700MQ, if I went much below 1.80 V, it would crash. When I started to increase the VCCIN beyond 1.85 V, the reported power consumption decreased. Based on this testing, I am pretty sure that the default VCCIN is somewhere between 1.75 V and 1.85 V. My best guess is about 1.80 V but I don't know if it is possible to come up with an exact number.

 

Well tried again to request some data for updating limit reasons for SkyLake but after 140 views, 3 downloads and zero replies there's is no interest in this. Fair enough I guess, less work for me and maybe you too.

Strange you mention power limiting on these 15W processors as I also wrote some software to remove PL1, PL2 and PL3 limits but once again couldn't get anyone to test and provide some feedback. For some weird reason I thought people with Haswell and Broadwell low power processors might be interested in that or even some of the quad users but seems not.

 

Will help from @Phoenix with his 6700K help you? I'm pretty sure he'll be happy to help

 

No time to test or change a single thing with my system after I have installed everything and everything working perfectly.. I have wasted enough time getting this damn throttling machine to be stable.

 

That is why I have not released a new version of RealTemp in 4 years. It became too painful begging users and waiting forever to get some feedback. Everyone is too busy so I just work on my freeware projects at my own pace. I appreciate the users that do offer to help but I try to only add new features for hardware that I have access to. When trying to write some code, I cannot wait days or weeks for someone, somewhere to do a simple 1 minute test. No Skylake CPU in my house at the moment so improving Skylake support will sit on the things to do list until I get around to buying a 6600K. Maybe next month or next year or maybe never. No hurries.

@Phoenix - The Eurocom Sky X9 looks like a beast. I might have to ask Eurocom if they would approve of a special TS version to celebrate the new laptop king.

http://i.imgur.com/rILokUl.png