The ThrottleStop Guide

@plee82 - If the iGPU is disabled by the bios and not being used, that might explain why there is no power consumption reporting for it. Did you try monitoring with HWMonitor as well?

@Eason - Just curious, did you try adjusting both CPU Core and CPU Cache equally? Did you also try switching those 2 both from Adaptive to a fixed Static voltage. It looks like it should work so not sure what is holding it back. Did you also try adjusting all of the other voltages a little bit to see if any of them work?

That wouldn't be a bad processor with it having +6 bins of turbo overclocking available but I see that Apple has Locked that feature too. Under volting makes such a big difference on Kaby Lake. It will be a sad day when all manufacturers disable this feature.

 

Hmmm core 0 and core 2 are hotter than 1 and 3 by average of 5 degrees at max temperature. Ugh I do not wanna repaste...


Sent from my iPhone using Tapatalk

 

@plee82 - Intel's core temperature sensors are only accurate to +/- 5°C so I definitely wouldn't worry about that.

 

Oh wow good to know. I did prime95 to test my undervolt and core 0 and 2 were hotter like 7 to 9 degrees more than 1 and 3...

 

The only time you need to worry about core temperature is if it is interfering with achieving maximum performance. During normal use, whether your cores are running at 60°C, 70°C, 80°C or 90°C really doesn't make much of a difference. All of these temperatures are consider to be "safe" by Intel. Your CPU should be able to have a long and happy life at any of these temperatures. If you have a K series CPU and you are trying to overclock the last possible MHz out of it then applying the best paste money can buy makes sense. For most other situations, why bother. Your games will never know that your core temps are not perfectly in sync.

 

I tried it all but nothing seemed to stick. US$3300 computer that's getting returned tomorrow. Too bad for Apple.

 

Thanks. I started off there and went up to about -125mv and now at -150mv and everything seems stable. I tried going up to -200mv or so earlier and got a BSOD about "Clock Watchdog Timeout" which I can only assume had to do with giving it too little power so I am going to sit at -150mv for now and do some more testing and benchmarks and try gaming and if everything seems stable I will try to increase the undervolt in smaller increments from here now. I had never used Throttlestop at all before so I was a little nervous to start messing with too many things and wasn't sure exactly what I had to change at first but it seems fairly simple for this undervolting and that there's only 2 or maybe 3 options to really change or adjust, and I was a little nervous about going too low but it seems like I did that once and it didn't kill me.

Now, in this laptop the iGPU is completely disabled and not an option to be used, would I still need to go into the iGPU in Throttlestop and change anything, is that still drawing and using power and potentially adding heat to the machine if it's disabled and not in use? I know it may be a stupid question but I want to make sure if I should be doing anything there or leaving it alone if I am not using the iGPU at all.

 

Core difference of 5C is okay. But 10C or more means you need to repaste your AW 15/17.

 

Wise choice. Maybe invest in another Thinkpad or MS Surface or even Dell Inspiron.

 

Playing games all cores are close. But in prime95, core 0 and 2 are 5-8 degrees hotter.


Sent from my iPhone using Tapatalk

 

Don't use p95 it will burn out your cpu and board.

 

Is it that bad? I was not getting thermal throttle but TDP throttle a lot.


Sent from my iPhone using Tapatalk

 

Yes its bad. Ask @Mobius 1

 

Wanted to ask, my PROCHOT is 88 according to TS. Isn't that very low?


Sent from my iPhone using Tapatalk

 

Omg what happened? Do you have a link?


Sent from my iPhone using Tapatalk

 

Yes its low, what's your cpu model?

 

I7-7700HQ.


Sent from my iPhone using Tapatalk

 

AW 13?

 

MSI GT62VR


Sent from my iPhone using Tapatalk

 

That processor should never get close to 88C, even on automatic fans. It's locked to 3.4 ghz on all cores and can never exceed 45W without PL being triggered.

100% of 7700HQ's will undervolt by 100mv. Someone will go as low as 125-150mv (150mv will require careful stability testing). If you can manage 150mv, your cpu temps won't go higher than 75C even with automatic fans).

 

I am currently undervolted to -140mv. It is indeed not reaching 75 at all unless I run prime95 where it hits 85C. Btw, it is locked to 38 for turbo boost multiplier.


Sent from my iPhone using Tapatalk

 

That's for ONE CORE. Prime95 doesn't use one core unless you force it to. The 7700hq can NOT use 3.8 ghz on all 4 cores no matter what you attempt. And it's not a good idea to use prime95, IMO. Something much more reasonable, realistic and -real world- is Stockfish chess engine (and you can set the # of threads).

 

Yup that is true. Prime was actually TDP throttling a lot instead of even reaching thermal throttling.


Sent from my iPhone using Tapatalk

 

You can use prime blend. Blend shouldn't exceed 45W if you are at -140mv. I tested it before I returned my 7700HQ for a 7820HK system.

 

The default PROCHOT temperature for the majority of Intel CPUs is 100°C. Manufacturers that decide to use a lower PROCHOT temperature to protect their laptops do not have a lot of confidence in their products. This might be a warning sign that there is some under designed component that they are trying to protect. Intel CPUs are very robust and are rated to run reliably right up to the PROCHOT temperature. Intel calls this temperature the "maximum safe operating temperature". Some laptop manufacturers need to read the Intel docs.

 

Not too concern atm. All my games are running so cool it feels like I am playing on my desktop lol.


Sent from my iPhone using Tapatalk

 

Has someone epxerienced higher voltages than set in Throttlestop?
(The problem also occurs when setting the same values in XTU...)
I am trying to limit the maximum voltage that my device (an Alienware 15R3 @ bios 1.17) will supply its i7 6820HK.
However, regardless of my settings, the CPU will always show a maximum voltage of around 1.090 V...


Has anybody experienced something similar?
Or am I setting the wrong values?

Thanks a bunch for any advice

 

I see nothing wrong. Your voltage ID is EQUAL To your CURRENT voltage. Don't look at maximum voltage. Look at current.

You are also NOT using a fixed multiplier for all cores. The voltages will only be identical if the multiplier is the same for 1-4 cores. The multiplier is adaptive, not STATIC. That means as the CPU multiplier goes higher, the voltage ID will also increase (that is what adaptive means). And since you have different core ratios for 1-4 cores, the fewer cores are in use, the higher the voltage ID since the multiplier for the remaining core(s) will be higher.

 

@FrozenLord - In the FIVR window, did you try setting the Voltage slider to Default and then increasing the amount of Offset voltage to control the maximum voltage that way? The default voltage curve on many of the 6th and 7th Gen CPUs is a mess. You might not be able to use software to come up with the perfect voltage that you are looking for. You can end up with either not enough voltage when lightly loaded so you see light load BSOD crashes, or you have your problem with the CPU getting too much voltage when fully loaded, especially when running software with heavy AVX instructions.

 

After I uncheck BD PROCHOT, It works, my multiplier stay in 18 without charger plugged in..
I already switch it to High Performance too.

But after 2-5 minutes, my laptop accidently shutdown... no windows error report.
Maybe my battery can't hold that multiplier without charger plugged in....

 

Set XTU service to manual in services.msc, If its running Stop it.

 

The battery shut down to prevent damage to itself, the computer and to any humans nearby.

 

Kinda confused with Non Turbo Ratio. Mine right now shows 27 on my i7-7700HQ. This CPU is supposed to have a 2.8Ghz clock with turbo boost up to 3.8Ghz. Why is the Non Turbo Ratio at 27 instead of 28?

 

Me too. Different manufacturers set the Non Turbo Ratio up differently. The 3 most common default values that I have seen NTR set to is either 0, the default multiplier of the CPU or it is set to 1 less than the default multiplier.

If Speed Shift is disabled or does not exist, NTR = 0 allows the CPU to fully use the SpeedStep function so the ThrottleStop Set Multiplier feature and Windows will allow you to access pretty much any multiplier from the minimum to the maximum. Most users that complain that the ThrottleStop Set Multiplier feature no longer works need to set NTR to 0 before complaining.

The 7700HQ has a base multiplier of 28. If NTR is set to 28 and Windows asks for the 28 multiplier or less, it should get that multiplier. If Windows asks for any multiplier higher than 28, it should automatically go to full turbo boost and use the maximum multiplier if possible.

If your 7700HQ has NTR set to 27, a request for the 28 multiplier will send the CPU into full turbo boost. A slight difference between the last two options. You can set NTR to whatever value you like and you can combine this with the Disable Turbo function. The best way to understand this setting better is to put a light load on your CPU like a single thread of the built in TS Bench and then adjust NTR and adjust the Set Multiplier value, etc. and see what multipliers your CPU runs at. Some simple hands on testing like this is the best way to learn more about Intel CPUs. They have a lot of multiplier control options that go back a lot of years and these options often times overlap each other. Some newer options like Speed Shift super cede older options like SpeedStep and the Non Turbo Ratio can be used to fine tune things further.

On my 4700MQ, it does not have Speed Shift available so I set NTR = 1. What this does is that any time Windows requests a multiplier higher than 1, which is all of the time, the CPU will use the maximum turbo multiplier instead and will ignore all of the intermediate multipliers. Full speed ahead makes me happy.

If I had a 6th or 7th Gen mobile CPU, and I wanted to see a lower multiplier while lightly loaded, I would set NTR = 0. Then I would enable the Speed Shift function and allow the CPU to manage its speed internally. Testing at Anandtech showed that a modern Intel CPU can manage its speed better than Windows can manage this.

That is pretty much the problem. Using ThrottleStop to try and force a battery to deliver more power than it is capable of delivering is asking for trouble. Maybe you can try doing this again on the 4th of July. When your battery goes ka-boom, it might create a fireworks like display. Our American friends will be happy!

 

If you set the NPR to some number like 27 and disable Turbo Boost, what happens? Also, Do I have to check the Lock checkbox next to the NPR so that it takes effect?


Sent from my iPhone using Tapatalk

 

There is rarely ever a need to use the Non Turbo Ratio Lock function so avoid selecting that option. Just change the value and then push the Apply button. The Title of this option will show you what the current value in the CPU is set to. If the title says Non Turbo Ratio - 27 then the Non Turbo Ratio is currently set to 27.

It is N TR not NPR. Go do some testing. Change values, play with Turbo Boost and watch what happens to the CPU multiplier. I do not own a 7700HQ so make some adjustments and see what happens. Go be a scientist and do some tests. Post some screenshots if you need help understanding anything. I don't mind helping out but I would rather not do someone's homework for them.

 

Oh sorry I thought you already knew what would happen lol. Tested right now and as expected, the cpu maxes at 2.8 GHZ.

 

I read over 100 pages about the Speed Shift feature in TS today. It is just my impression but it sounds to me that if I want my cpu freq to be always at maxed, it is better to use the High Performance profile in Windows instead of Speed Shift. However, If I wanted to opt for something like the Balanced Profile in Windows, it is better to use Speed Shift. Are these assumptions correct?

 

Instead of all that reading, why not just enable Speed Shift, adjust it and see what it does?

For maximum performance, you can use the Windows High Performance power profile with the Minimum processor state set to 100%. You could also try setting the Non Turbo Ratio to 1 or you could enable Speed Shift and set the Energy Performance Preference (EPP) to 0. Those are the basics. Which method works best can only be determined by doing some benchmark testing. Different laptop manufacturers have dreamed up a lot of different throttling methods so what works on one laptop model might not work very well on a different model.

ThrottleStop is very efficient but when I use it, I make sure to check off as little as possible. I only check off what I need to fix the problem. Some users prefer to check off items randomly based on a screenshot they saw somewhere on the internet. That's why I always recommend doing some testing. There is no need to check off things like Clock Modulation unless your laptop is actually using this throttling method. For my Y510P, I can fix its throttling issues while ThrottleStop is in Monitoring mode. This combined with the Stop Monitoring feature saves as many CPU cycles as possible for more important things.

 

Yup been playing with that. The windows high performance profile already has the minimum state set to 100% by default.

By the way, tried Speed Shift with EPP set to 0. I am noticing turbo boost is not as aggressive as the Windows High Performance mode. During single thread bench, I rarely hit 38x. It is actually hitting 37x, I don't think I have ever hit 37 this often lol. When I try the Windows High Performance mode, the turbo boost much more aggressive.

Sent from my iPhone using Tapatalk

 

I mentioned that because some users claim to be using the High Performance power profile and then later on mention that they changed the Minimum processor state to 5%. In this case, they have basically converted the high performance profile into the balanced profile.

Are you watching this in HWiNFO or with some other monitoring tool? TS will never show the full 38.00 multiplier because there are always WIndows background tasks running which need to be processed which constantly interfere with the multiplier staying at 38.00 for an entire benchmark. Getting consistent and repeatable results when testing on a Windows 10 PC is a challenge. Some laptops have a lot of junk running in the background. The max multiplier on my 4700MQ is 36. During a single thread TS Bench test, I can get within a hair of the full 36 multiplier. Not a chance getting anywhere close to this multi with XTU running in the system tray.

 

TS shows 37.96x with the single thread test. With speed shift, i am more in the 36.90x and 37.00x. Does Non Turbo Ratio affect Speed Shift? I left it at default 27.

Maybe it is a placebo though, things feel a bit snappier.

Another interesting thing, now a different core is hotter than core 0!!! I actually have better heat distribution in my cores wow.

Sent from my iPhone using Tapatalk

 

I am noticing my Package Power Long is defaulting to 200, same with Power Limit Short. The clamp checkbox is marked in Package power Long. The 200 though is being ignored. I am noticing TDP throttle kicks right away at 45W. @unclewebb if the TDP throttle kicks right away at 45W, what is the point of the 200? I feel like the power limit is locked to 45W.

 

That is something that I have never heard of before. When testing, use something like Prime 95 Small FFTs. Most benchmarks are not very good at evenly loading the cores.

Is EPP on the main screen set to 0? When using Speed Shift, did you try setting the Min and Max both to 38?

This is a common problem with non K CPUs. There are multiple locations where the power limit can be set and ThrottleStop only has access to one of these limits. You can always load up XTU to see if it lets you go beyond the long term 45 Watt TDP limit. Probably not.

Maybe. I have not tested that yet.

Did you do some wPrime or similar testing yet? You can watch the TS multi while a single thread of this bench is running. That will keep you busy for a little while.

 

Exclude p95 because @Mobius 1 said its bad. Try AIDA64 or Wprime 32m or 1024m

 

occt for testing long term heat

realbench for cpu stress test

 

Yes I tried Prime95. I have my EPP set to 0. I am trying to figure out if there are any differences between the Windows High Performance profile and Speed Shift with EPP 0.

 

I tested this on a desktop Kaby Lake CPU using one thread of the TS Bench and the multiplier was exactly the same. When you said your multi wasn't going as high, it sounded like EPP was not set to 0. Double check this in the FIVR monitoring panel. It is also possible that the mobile CPUs work a little differently when it comes to EPP. Technically, this setting is referred to as a Preference. The CPU can decide to ignore this user preference and run however it likes.

I still like Prime95. The Small FFT test is one of the very few tests that equally loads the cores so it is useful when temperature testing. I used to be afraid of running P95 on my laptop because it is overkill and is not a realistic real world kind of load. The Small FFT test sends temperatures sky high.

Is this a bad thing? Maybe but I don't think so. The 7700HQ is rated at 45 Watts long term. The exact same 14nm technology used in the 7700K is rated to run reliably up to 91 Watts. Running Prime95 on a 7700HQ is not going to hurt the CPU. It will throttle down to the 45 Watt limit if it needs to which is perfectly safe for this technology. If the CPU gets too hot, it will throttle to keep the core temperatures below 100°C which Intel lists as the "maximum safe operating temperature".

The next argument against Prime95 is that it will hurt the motherboard. Why? The motherboard manufacturer knows that this motherboard must be able to handle running a 45 Watt CPU. The power adapter, etc., all must be designed to meet this requirement. If anything fails during this sort of stress test then that is a sign of a defective component or poor design. The built in turbo power limits and core temperature sensors both force you to run the CPU within spec. If a laptop cannot handle running within spec, send it back to the manufacturer and tell them they can keep it.

If you are still scared of Prime95, use ThrottleStop to run Prime95 at a lower multiplier.

 

So been testing more and noticed it does go as high but it does not stay as much that high. Gaming wise there are no difference. I did read a thread where EPP 0 had better real time performance doing some latency dependent stuff with music.

 

Why is it that my set-up is always going to profile 1 even if I choose another profile to run? Is the program set this way? When I choose profile 3, after sometime it will automatically revert back to prof 1.

My set-up is 1-performance, 2-fail-safe, 3-internet/entertainment, 4-battery. All four have the same undervolt settings, profile 1 running at its max turbo freq.(4.5) while having BD, CState disabled and Speedshift is set to 0. Profile 3 is set to stock frequency 4.2, speedshift is 128 and nothing is disabled.