The ThrottleStop Guide

@Takeishi

i never use my laptop unplugged.. it's a big 17' desktop replacement.. sometimes i leave my pc on ALL night with my fav games running just because i like to leave my thing online.. so i mean maybe i should set it to 80? or if i'm always plugged i can regardless always use the 0 -3 -4 profile without fearing for the longevity of my cpu?

 

I'm pretty sure this limit is dependent of what the manufacture has set, and not on your version of throttlestop (ex. mine 95 C). You could always try unticking the check box, but it might be impossible to change it, especially if it's a laptop CPU.

I don't have a lot of experience with temperature controlled limits, so I'll let someone else take over.


Envoyé de mon iPhone en utilisant Tapatalk

 

In this situation, with a laptop, I would let it at around 6+, this is mainly personal preference tho. I just don't like to leave it at max for nothing, and also my laptop's CPU and GPU use the same heat pipe so letting the CPU run at max turbo when gaming would heat up my GPU needlessly.

 

256 used to be a magic number but you can try any big number. If you set it to 1024, your CPU is not going to have 1000 Amps flowing through it and instantly go ka-boom. You are just trying to see if the Current Limit setting is the reason for your throttling. Set this sky high in the bios and see if EDP in red goes away when stress testing. The AUTO setting might not allow enough current for your CPU to maintain its full overclocked speed when fully loaded with Prime95 / LinX.

Is that a bad thing? If your laptop is always plugged in, using a SST - EPP setting of 0 and letting the CPU run at its full Intel rated speed is not a problem. Many users with the unlocked K series CPUs are running this same technology at 5.0 GHz so running your i7-7700 at 4.0+ GHz is fine. Your original screenshot showed way too much background activity. Getting your C0% down to 1.0% when idle is where it should be. You do not have to go crazy like I do and pick through every Windows background task with suspicion in search of an uber low C0% number. I just like showing people what is possible. Windows 10 has some very inefficient code running in the background at times.

People worry too much about high MHz. If you have your C States enabled and individual cores are entering C7, the core in this state is disconnected from the voltage rail so it is getting 0 volts and its clock is stopped internally so it is running at 0 MHz. The only time a core should come out of this state is when it has something to do. Might as well let it run at full speed so it can get whatever it needs to do done quickly and then it can get back into the low power C7 state, the sooner the better.

@agungin - ThrottleStop cannot be used to change the PROCHOT temperature when your CPU will start to thermal throttle and slow down. This has to be taken care of before Windows starts so either in the bios by the manufacturer or with a UEFI shell command which is beyond me.

 

@unclewebb

i take a look to the C state window while my pc was semi idling just with couple apps opened( not high load) and this is the screenshoot

https://gyazo.com/d50698abce3eeb9d9d6ba1c27ee33b9d

Could you explain to me the meaning of the section core c state percent and package c state percent? to try to understand better the thought you just explained in your previous post.. thank you so very much!

 

UncleWebb can give you more details, but the gits of it is:
-Core c state percent (don't bother with the rest)
-your CPU has multiple C state, the higher its number, the lower the state is, and the less energy it spends.
-When doing work, the CPU will mostly be in the higher states, the highest (I believe) been C0. Your CPU will try to complete his tasks and go back in the low C7 state.
-The % is the percentage of the time the CPU is spending in each states. A C7 of 50% means that your CPU is spending half of its time in the lowest state (so you could say it's mostly active). Each columns represent one core of your CPU.
-My laptop will be mostly active when plugged and running at a high clock speed, but on the battery there will usually be 2 cores at 98%+ in the C7 state (so pretty much completely idle) one at 90%+ and one between 80-90%. This shows me that my laptop works efficiently on battery and doesn't waste time in higher states when there's no processes running.

But it's not really as much of a concern were you're plugged in.

Edit: for a plugged laptop running a couple programs at 4.1Ghz, your results looks fine to me.

Envoyé de mon iPhone en utilisant Tapatalk

 

@unclewebb
then it means i couldn't tweak or at least use throttlestop to stop the lag when i play a game or rendering before i figure out what my thermal issue?

is it common to have 88c full load on a laptop, or it's just mine which had an issue?

 

I get about 86-88c full load on my 6700hq with a -125mv undervolt. Before, it could go from 90c to its max of 95c in intense gaming sessions. What kind of laptop is it?

Edit: temp limits can be hard to bypass especially on laptop. I would like bump my GPU's from its 82c limit, but if it's not accessible in the BIOS then it's pretty much hopeless.


Envoyé de mon iPhone en utilisant Tapatalk

 

I would say it is common on poorly designed laptops that use heatsinks and fans that are inadequate. A 3rd Generation Core i3 does not put out a lot of heat. It would be best to open up your laptop and blow out any dust from the heatsink and fan. If that doesn't help, you could try cleaning and replacing the thermal paste between the heatsink and CPU. If that doesn't help, time for a laptop from a different manufacturer. The world wants small and quiet laptops that also have to be inexpensive so high core temperatures are the result. Some manufacturers have been known to use the thermal throttling temperature as a way to save a few pennies so they don't have to spend money and include a proper heatsink. Shameful but all too common.

 

speaking of high temp, what is a good temperature limit you would consider on a well built laptop? my laptop is fresh new , it's a gaming one from a reseller that assured me it was perfect for the cooling thing ( double fan with separated cpu and gpu etc ) , they claim to be so good that they offer 3 years warranty without extra charge. back to the point, when i first got the laptop 2 weeks ago i would reach in intense gaming sessions with heavy games like bf1 or ghost recon wildlands medium temperature around 84 with peak to 90.. i called support and they said it was perfectly fine. I undervolted then to -120 using throttlesop and now i have an average temp of 78-79 with peak to 84 : is that ok? i'm referring to gaming sessions... if i use just benchmarks for cpu that just super load the CPU , it wont even get over 70 degrees.. but apparently games are way more stressful than those bencharmks.
and on other thing, when i first tried to find my sweet spot to undervolt i downed it all the way to -140.. and during a game session i had like a reboot of my pc.. a crash of the machine.. so i lowered it to -120.. i've not been having crashes for 1 week and then today suddenly i have a crash.. so now i lowered it to 110. Is this the common way to proceed.. right? i mean i don't explain why in one week i never had an issue and suddenly today it happens but i guess this is the way to find the spot for the undervolt because it depends from many variables that maybe every day are not tested each in a single way? I hope i explained my self, sorry about the confusion :/

thank you!

 

Even with 2 separate fans, the heat from the GPU will affect the overall temp of the laptop during a gaming session, which will up the CPU temp a bit. In a CPU load test, only the CPU is running.

Which test did you use for the undervolt? I personally use prime95 blend test 2h and the unigine DX11 Heaven Benchmark at ultra and 1080p for 2h (there are other alternatives, p95 makes the CPU heat up quite a bit).

I did not have problem when gaming, but even if your setup holds during a test, it doesn't mean it will always be stable enough. So if it crashes during the following days/weeks (just like in your case) just lower it and see how it goes.


Envoyé de mon iPhone en utilisant Tapatalk

 

ThrottleStop 8.48
https://www.sendspace.com/file/hcmblz



New Features
- special OBSIDIAN-PC edition.
- enable Kaby Lake Limit Reasons reporting.
- improved Skylake and Kaby Lake U CPU BCLK and multiplier reporting.
- previous Start/Stop Monitoring status applied at start up.

 

submitting it to Major Geeks now

 

That didn't take long!

A few months ago, my friend John over at Obsidian PC asked me if he could include a version of ThrottleStop with his Obsidian Tool. He was nice enough to ask so I decided to make a special version for him. Every time I showed him my progress, he countered with another suggestion. It ended up taking far longer than I anticipated but I like it and hopefully John likes it too. I am always open to suggestions so if you see something you like or don't like, let me know. This version is a work in progress.

 

as I've always mentioned, I really wish there could be a way to automate the startup of TS automatically rather than having the user to manually create the task

 

There you go Mr. Unclewebb, updated it for you on MajorGeeks with new image for the new GUI: http://www.majorgeeks.com/files/details/throttlestop.html

It has been downloaded 9574 times so far

 

I have always avoided adding a "Start with Windows" feature to ThrottleStop to save the rookies from hurting themselves. Some people set ridiculous voltages, do not bother doing any testing and then end up with an unstable laptop. It is too easy to get stuck in an infinite boot loop if Windows is trying to start up with inadequate voltage. I need to add in some protection to prevent this before adding ThrottleStop to the Task Scheduler automatically. Maybe someday.

@crusher88 - When testing, try to find a game demo that loops so you can run it for a few hours to see if your computer is stable or not. If an active GPU is adding heat to your laptop and to your CPU then doing CPU only stress testing doesn't go far enough. As CPUs heat up, they tend to become less stable so voltage settings that work great at a core temperature of 75°C might not be stable at 90°C.

 

i used the same tests you use but the games i've been playing are very " punishing " .. ghost recon wildlands just came out and it stresses a lot my pc so i think that's the best benchmark i could actually do.. only problem is that i would need to do a game session of 5 hours non stoop XD

 

Yeah, I made that mistake with afterburner. Checked the run at startup box, but ended doing some more testing and crashed it so it definitely can happen :/

It uses "start on login" tho, not start with Windows (at least with Win 10), and that's what I use with ThrottleStop. If it crashes, Shift + Reboot button on the login screen to get in safe mode and you're out of trouble.




Envoyé de mon iPhone en utilisant Tapatalk

 

So @unclewebb I did some moar testing, and I determined that while it's a very different scenario to what my 4800MQ was, I can get it working for sure.

Seems I need to match my cache and core voltages in throttlestop or undervolting won't work beyond what I set in the BIOS.

Manual voltage works, but only with "Static" and not with "Adaptive". I'll do some more checking to see if this is some kind of isolated behaviour with me, or something to that effect.

 

Android voltage control apps have a feature where they start on boot but with a one minute delay (or the program could start but only apply settings after a one minute delay) so the program can easily be closed before voltages are applied.

 

@crusher88 , you can use OCCT 4.4.2, running 3D test with clicking check error box as well as Prime95 v28.10[First one, Small FFTs] for your testing purpose. In my laptop, it can go up to 31W(~19W[CPU]+~12W[GPU]) with i5-4200U CPU. The temperature is up to 88 degree.

 

@unclewebb Any idea if it's possible to upgrade this laptop (HP ProBook 6460b) to the chip you mentioned (i7-2720qm)? The official literature seems to indicate that I've got the most powerful i7 (dual core) for this laptop, but it uses the same chipset and cooling solution as the Elitebook 8460p, which can take quads.

 

Check this thread:
https://forums.overclockers.co.uk/threads/hp-probook-6460b-cpu-upgrade.18699997/

Compatible list for Intel HM65 chipset:
list of compatible CPU's for the chipset
http://www.cpu-upgrade.com/mb-Intel_(chipsets)/HM65_Express.html


Also (this one is for the 6560b):
h30434.www3.hp.com/t5/Notebook-Hardware-and-Upgrade-Questions/Hp-probook-6560b-Upgrade-processor-help/td-p/2257171

Interesting answers from the thread:

"I installed a i7-2620M which worked fantastically, I did some research and asked a couple of intel guys and they said I can stick any 2nd gen i7 in the probook so I am now running the i7-2820QM in a probook 6560b with 16gb ram, you really notice the difference with 4 cores even though the 2620 had a turbo it's still only on one core."

Followed by:

"running the i7-2820QM flawlessly, 1 hour stress test got max 74 degress across all four cores, running at idle is 30 - 40, hope it helps guys"



Envoyé de mon iPhone en utilisant Tapatalk

 

I can only say uncleweb is one of the coolest guys i had the pleasure to meet in this forum!
It´s been a pleasure to supply him with input and it´s a pleasure for our Brand to be able to have this TS version for our clients.

 

and its nice that you share your Obsidian tool with the comminity too - sadly you do not do so with the BIOS :-( i would even like to donate for it... (but i understand the reasons)

 

If you want to donate to someone please do it, donate to uncleweb.
As for the BIOS it would raise havoc with other brands/resellers, im sorry i cant do that...

 

Hello, I consider buying a Clevo N650DU or W650KK1 and putting i7-7700k in it. There are some thermal challenges with it and, as a part of the solution, I'd like to undervolt the CPU. I've been told by a reseller than it's not possible in BIOS, so OS is the only way.
1. Do you have a suspicion that it may be blocked by the motherboard (h170)?
2. I'd like to lower it with TS, but I can't because I'm not going to put Windows on that machine. Linux and FreeBSD for me. I know what are MSRs and how to set them, but would need some assistance as to where can I find voltage offset. I looked up Intel's manual on Kaby Lake MSRs, but can't find this one.
unclewebb, could you help me with that?

 

my reseller told me that his N650DU does not have an Bios Update for Kaby Lake microcodes at the moment. And he has no time schedule if this ever will happen....

 

Some offer it already:
https://www.mysn.de/schenker-office-laptops/schenker-office15
https://www.multicom.no/systemconfigurator.aspx?q=st:10636518;c:100561;fl:0
https://www.tuxedocomputers.com/Lin.../SSD-bis-32GB-RAM-DVD-o.-Blu-Ray-Brenner.geek

 

Hello mates. I'm playing with the new toy that those kind guys at Obsidian send me. It's a P650 with a 7820HK.

Now, of course i want to have some fun and get the best out of it. I could "easily" get it to 4.4GHz with hyperthreading on and now i was playing around at 4.5-4.6 Ghz with hyperthreading off. When running cpu intensive task it's not really a huge problem. Temps go up but can still manage to get those clock with a stable system.

All of the "issues" come out with benchmark or real life softwares that stress the FPU. I'm now at 4.2Ghz overclock and if i test the CPU with AIDA64 it's all ok, cpu stays at 4.2GHz and northbridge stays at 3.6GHz (actually this is a little weird since inside the FIVR mask of ThrottleStop i've set cache ratio up to a maximum of 40x multiplier but it never goes beyond 36x multiplier unless i force the cache multiplier at a fixed ratio and then it actually goes over 36x with no problems).
The instant i switch from CPU ONLY test to FPU ONLY test both the cpu and northbridge throttle down a little. CPU goes at 3.6GHz and northbridge at 3.2-3.1GHz. I can get the same weird behaviour even with x264 encoding or real life finite element simulations since all these softwares stress the FPU.

I thought it was some sort of power limit throttling (it's not thermal throttling for sure) but i've set power limits 1 and 2 up to 100W from bios and even power limit timer to the maximum of 128s. Moreover I've set everything i could in the throttlestop TPL mask to the maximum value to be sure there would have been no power limit affecting the cpu. In fact when this kind of throttling appears i can see from TS that EDP is limiting both CPU and RING and in HWInfo it shows that the limits are regarding "IA: Electrical Design Point/Other (ICCmax,PL4,SVID,DDR RAPL)" and "RING: Max VR Voltage, ICCmax, PL4".

At this point i was like: ok the cpu is not getting enough juice and Icc max is too low. I went to the bios and changed Icc max through the VR setting feature. I went up to 250A and nothing changed. Same throttling, same reasons. I really have no clue of what's happening.

Also the Whole thing doesn't look very dependant on the OC i'm applying. At 4.4Ghz and 120mV higher voltage offset the same happens. CPU stress, everything stays at max clock (cpu 4.4 and northbridge at 3.6) while on FPU stress everything throttles pretty much at the same clocks as before (cpu 3.5-3.6 and northbridge at 3.1-3.2).



PS: MANY MANY thanks to unclewebb for his precious work!

 

Ok. I can confirm it's something related with Icc max. I installed XTU under the advice of Obsidian and forcing higher Icc max from XTU works. I could stress FPU at 4.2GHz with no throttling (although temps might go crazy) at all.

Now, i don't want to use XTU because throttlestop is so much better plus XTU might mess up fan tables (although i'm not reporting any problems right now). The problem of not using XTU arises because it looks like i cannot touch Icc max from throttlestop and the BIOS settings looks like it doesn't do anything in reality.

Any good idea to find a workaround?

 

Prior to the Creators Update the SST EPP settings would stick. However today after installing the update I have noticed that Windows will change the SST EPP depending on the Power Plan when I switch windows etc. However if no window movement is detected, TS seems to take over and apply whatever has been configured.

Here is what I have found out:

• High Performance = SST EPP 0
• Balanced = SST EPP 84
• Battery Saver = SST EPP 128

A) Does anyone notice the same behaviour?

B) Any workarounds?

 

Do you see it changing inside TS? Because on mine it sticks to what is set. At least that's what I see from inside TS.

 

Yup, I am using the FIVR panel to monitor it.

It switches like when I change windows or move them about. Sorry, not the mouse. It changes also like during games.

 

I was monitoring it from main panel (not inside FIVR) and in my case it never changed.

 

Still, on my laptop it's fine. Win10 is updated and TS is too.

 

Are you installing the preview release of Redstone 2 15063.3? Because I'm on that and I see no difference when dualbooting with 14393,. SST doesn't change for me.

But I'm on power saver all the time.

 

Ok so after some research and some testing on my computer, I have found that in Windows 10, you can adjust hidden power options in power plans to get the Speed Shift to stick ( https://msdn.microsoft.com/en-us/library/windows/hardware/mt422955(v=vs.85).aspx)

The PerfEnergyPreference directly controls the OS input to the SST register, which overides Throttlestop when the SST requested is lower then the one in TS.

Here is the descrption from Microsoft:

Code:

When set to 0, the energy performance preference register is programmed to 0 to favor performance. When set to 100, the energy performance preference register is set to 255 to favor energy savings. When set to an intermediate value, the energy performance preference register is programmed to the value: (setting * 255) / 100.

When I check for my High Performance plan, the value was set as '0' so the OS would change the SST register to ) whenever it detected that more power was needed.

 

I don't think I saw you mention it, but was your windows power setting set to Performance (not Balanced or others) and min/max CPU usage both set at 100%?

Edit: I believe power managing of the processor was given to the CPU since Skylake and newer, and shouldn't (or doesn't need to) be managed by the OS anymore. I saw that somewhere in the literature on SST linked by UncleWebb. Power setting other than the one mentioned could cause the OS to want to take back control over management, from what I understand.

Envoyé de mon iPhone en utilisant Tapatalk

 

The main panel does not report the current SST settings, rather it is what you are asking TS to set for you. The SST-EPP number in the FIVR panel is the reading from the processor register

 

The power plan is High Performane, min 1% max 100%.

Yes, the power mangement is handed over to the processor if the PerfAutonomousMode or Processor Performance Autonomous Mode is enabled, which by default is.

After a little more testing it seems to work as follows:

If Autonomous Mode is enabled, it will default to the SST set in Energy Policy ( PerfEnergyPreference) when the Latency Sensitivity Hint ( PerfLatencyHint) is triggered, which under High Performance plan is set at 99%. When idle, the TS setting will then apply.

PS I've embedded the links to the relevant documentation for easy reading!

 

Nope, I am not on any insider build/ring/beta program.

 

Hey guys got a quick question for you. Every time I start up my laptop and check the limits reasons the TS PL1 and EDP other is always yellow. I am running the latest version 8.48 I think it is. Any ideas and/or suggestions?
http://i.imgur.com/dtlMx40.jpg
Thx

 

I will try again when I have some time then. Although I think I also tried to open FIVR.

Also i found out that fixed cache ratio multiplier is bugged and applies to every profile even when set differently. Does Unclewebb still follow o should we report bugs directly through PM?

 

Hi guys!
I need help!
I was using ThrottleStop v8.40 with no problems whatsoever, i updated ThrottleStop to v8.48, the settings sticked from the last version to the new one, but now nothing shows on my taskbar (the TS icon and i used the CPU Temp option) even though i checked that option.
I never had this problem before when i updated TS to new versions, and i don't know what happened this time...
I have a laptop MSI GS73VR-7RF with Windows 10 Pro 14393.969 fully updated.
Hope someone could help me with this nuisance, since this is more a nuisance than a big issue.

 

Thanks leader288,

I find the Intel SpeedShift (HWP) options to be numerous and complex. To confuse matters, on my Dell XPS 9550, some options can be changed but somewhere the computer just ignores the settings. I have difficulty evaluating some of the obscure settings if I don't even have confidence the changes are "sticking".

Consequently, I have assumed for max performance on my laptop:

- running Windows Power Options at "high performance"

- running SpeedShift with "EPP=0" gets me close to max performance

- setting Dell Command-Power Manager to "Ultra Performance" (at least for fast fans

- ThrottleStop c1e (check the box)

==> This is just trial and error and is probably not 100% optimal. But latency is reallly quite low now

==> (FYI, on my laptop, Dell requires C-States and SpeedStep to be enabled or Turbo CPU speeds get locked out. SpeedStep seems to be overridden by SpeedShift so that does not seem to matter. ThrottleStop has the fantastic C1E button that kills latency problems.)
_____
For those who want to read more, the linked Intel guide p.473 forward (HWP section) provides some helpful doucmentation although is a bit dense.

p.480 are tuning recommendations for max performance (or max energy savings)

http://www.intel.com/content/dam/ww...eveloper-system-programming-manual-325384.pdf

Uncle Webb helped us with SpeedShift on the Dell XPS there are a few ways to enable it. ThrottleStop is the easiest as it just requires checking a box.

http://forum.notebookreview.com/threads/dell-xps-speed-shift.796891/

 

256A+ should be the best way to kill that kind of throttle from the current limit perspective, however I don't know why TS and your BIOS doesn't work but XTU will. That is weird. XTU does interface with the BIOS unlike throttlestop as far as I know, so... that might explain the change.

OBSIDIAN-PC's installation of clevo's control center breaks XTU's access to the fan tables, so this is not a concern for you.

Also where are you attempting to touch ICCMax in Throttlestop?

That's fine, there is nothing wrong. Even my system has this with the 7700K set to 150W TDP and 256A current limit.

It's commonly stated you must run notification area cleaner.

 

D2 Ultima,

I searched and i was not able to find an answer to my problem! But your suggestion worked though!
Thank you for your help!

 

Sorry JerryDD, Intel voltage control info cannot legally be shared without signing a NDA with Intel.

@leader288 - If ThrottleStop and Windows are both trying to control the same CPU register, one of them is going to win and one will lose control. If you want ThrottleStop to control EPP then I would set EPP to 0 in Windows so it does not interfere with ThrottleStop. I will look into making ThrottleStop more aggressive when it comes to monitoring and setting the EPP value.

The default High Performance plan has the Minimum processor state set to 100%. If you have decided to change that value then the plan you have created is really not the High Performance profile anymore. I would use the default Windows High Performance power profile and then I would try setting up 2 profiles in ThrottleStop, one with EPP set to 0 and a second profile set to 80 or 128. That might let ThrottleStop control things without any interference from Windows. I have not updated to the latest version of Windows 10 so I am not sure how aggressive it is at managing EPP.

@Matthew Gary - Limit Reasons items in yellow mean that type of throttling was previously detected. It is not unusual or a concern if a few of these items are yellow after booting up or resuming from stand by. The CPU is in the process of being setup so anything is possible. Any item in red means that throttling is in progress. Now you need to be concerned. You can click on the headings in Limit Reasons to clear the yellow boxes so you have a clean slate before testing.

Holy crap. I have been around here for the over 6 years and have commented on the vast majority of user issues over the last 580 pages. Whenever I release a new version, it is always nice to have a day or two off.

When playing with the cache ratio, did you also change the Min and Max cache ratio? I will look into the bug you have found. I just did a quick test and found that the Min and Max cache ratio allowed me to control the cache ratio on different profiles without any problems. I am always interested in hearing about bugs as well as trying to fix them. It helps me a lot if you can post some screenshots so I can see what the problem is. I believe that Unlock Adjustable Voltage has to be checked for each profile that you want ThrottleStop to control the cache ratio.

On all of Intel's locked, non K processors, the C states like C3 or C6 or C7 must be enabled in order for the CPU to use maximum turbo boost. For once Dell might be innocent.